]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - net/xfrm/xfrm_policy.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: remove dst gc related code
[mirror_ubuntu-artful-kernel.git] / net / xfrm / xfrm_policy.c
CommitLineData
a716c119 1/*
1da177e4
LT		 2 * xfrm_policy.c
3 *
4 * Changes:
5 * Mitsuru KANDA @USAGI
6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8 * IPv6 support
9 * Kazunori MIYAZAWA @USAGI
10 * YOSHIFUJI Hideaki
11 * Split up af-specific portion
12 * Derek Atkins <derek@ihtfp.com> Add the post_input processor
df71837d 13 *
1da177e4
LT		 14 */
15
66cdb3ca 16#include <linux/err.h>
1da177e4
LT		 17#include <linux/slab.h>
18#include <linux/kmod.h>
19#include <linux/list.h>
20#include <linux/spinlock.h>
21#include <linux/workqueue.h>
22#include <linux/notifier.h>
23#include <linux/netdevice.h>
eb9c7ebe 24#include <linux/netfilter.h>
1da177e4 25#include <linux/module.h>
2518c7c2 26#include <linux/cache.h>
68277acc 27#include <linux/audit.h>
25ee3286 28#include <net/dst.h>
6ce74ec7 29#include <net/flow.h>
1da177e4
LT		 30#include <net/xfrm.h>
31#include <net/ip.h>
558f82ef
MN		 32#ifdef CONFIG_XFRM_STATISTICS
33#include <net/snmp.h>
34#endif
1da177e4 35
44e36b42
DM		 36#include "xfrm_hash.h"
37
a0073fe1
SK		 38#define XFRM_QUEUE_TMO_MIN ((unsigned)(HZ/10))
39#define XFRM_QUEUE_TMO_MAX ((unsigned)(60*HZ))
40#define XFRM_MAX_QUEUE_LEN 100
41
b8c203b2
SK		 42struct xfrm_flo {
43 struct dst_entry *dst_orig;
44 u8 flags;
45};
46
418a99ac 47static DEFINE_SPINLOCK(xfrm_policy_afinfo_lock);
37b10383 48static struct xfrm_policy_afinfo const __rcu *xfrm_policy_afinfo[AF_INET6 + 1]
418a99ac 49 __read_mostly;
1da177e4 50
e18b890b 51static struct kmem_cache *xfrm_dst_cache __read_mostly;
30846090 52static __read_mostly seqcount_t xfrm_policy_hash_generation;
1da177e4 53
25ee3286 54static void xfrm_init_pmtu(struct dst_entry *dst);
80c802f3 55static int stale_bundle(struct dst_entry *dst);
12fdb4d3 56static int xfrm_bundle_ok(struct xfrm_dst *xdst);
a0073fe1 57static void xfrm_policy_queue_process(unsigned long arg);
1da177e4 58
12bfa8bd 59static void __xfrm_policy_link(struct xfrm_policy *pol, int dir);
29fa0b30
WY		 60static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
61 int dir);
62
e37cc8ad
FW		 63static inline bool xfrm_pol_hold_rcu(struct xfrm_policy *policy)
64{
65 return atomic_inc_not_zero(&policy->refcnt);
66}
67
bc9b35ad 68static inline bool
200ce96e 69__xfrm4_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 70{
7e1dc7b6
DM		 71 const struct flowi4 *fl4 = &fl->u.ip4;
72
26bff940
AD		 73 return addr4_match(fl4->daddr, sel->daddr.a4, sel->prefixlen_d) &&
74 addr4_match(fl4->saddr, sel->saddr.a4, sel->prefixlen_s) &&
7e1dc7b6
DM		 75 !((xfrm_flowi_dport(fl, &fl4->uli) ^ sel->dport) & sel->dport_mask) &&
76 !((xfrm_flowi_sport(fl, &fl4->uli) ^ sel->sport) & sel->sport_mask) &&
77 (fl4->flowi4_proto == sel->proto || !sel->proto) &&
78 (fl4->flowi4_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 79}
80
bc9b35ad 81static inline bool
200ce96e 82__xfrm6_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 83{
7e1dc7b6
DM		 84 const struct flowi6 *fl6 = &fl->u.ip6;
85
86 return addr_match(&fl6->daddr, &sel->daddr, sel->prefixlen_d) &&
87 addr_match(&fl6->saddr, &sel->saddr, sel->prefixlen_s) &&
88 !((xfrm_flowi_dport(fl, &fl6->uli) ^ sel->dport) & sel->dport_mask) &&
89 !((xfrm_flowi_sport(fl, &fl6->uli) ^ sel->sport) & sel->sport_mask) &&
90 (fl6->flowi6_proto == sel->proto || !sel->proto) &&
91 (fl6->flowi6_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 92}
93
bc9b35ad
DM		 94bool xfrm_selector_match(const struct xfrm_selector *sel, const struct flowi *fl,
95 unsigned short family)
77681021
AM		 96{
97 switch (family) {
98 case AF_INET:
99 return __xfrm4_selector_match(sel, fl);
100 case AF_INET6:
101 return __xfrm6_selector_match(sel, fl);
102 }
bc9b35ad 103 return false;
77681021
AM		 104}
105
a2817d8b 106static const struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
ef8531b6 107{
a2817d8b 108 const struct xfrm_policy_afinfo *afinfo;
ef8531b6 109
a2817d8b 110 if (unlikely(family >= ARRAY_SIZE(xfrm_policy_afinfo)))
ef8531b6
ED		 111 return NULL;
112 rcu_read_lock();
113 afinfo = rcu_dereference(xfrm_policy_afinfo[family]);
114 if (unlikely(!afinfo))
115 rcu_read_unlock();
116 return afinfo;
117}
118
d77e38e6
SK		 119struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos, int oif,
120 const xfrm_address_t *saddr,
121 const xfrm_address_t *daddr,
122 int family)
9bb182a7 123{
37b10383 124 const struct xfrm_policy_afinfo *afinfo;
9bb182a7
YH		 125 struct dst_entry *dst;
126
127 afinfo = xfrm_policy_get_afinfo(family);
128 if (unlikely(afinfo == NULL))
129 return ERR_PTR(-EAFNOSUPPORT);
130
42a7b32b 131 dst = afinfo->dst_lookup(net, tos, oif, saddr, daddr);
9bb182a7 132
bdba9fe0 133 rcu_read_unlock();
9bb182a7
YH		 134
135 return dst;
136}
d77e38e6 137EXPORT_SYMBOL(__xfrm_dst_lookup);
9bb182a7 138
42a7b32b
DA		 139static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x,
140 int tos, int oif,
9bb182a7
YH		 141 xfrm_address_t *prev_saddr,
142 xfrm_address_t *prev_daddr,
25ee3286 143 int family)
1da177e4 144{
c5b3cf46 145 struct net *net = xs_net(x);
66cdb3ca
HX		 146 xfrm_address_t *saddr = &x->props.saddr;
147 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 148 struct dst_entry *dst;
149
9bb182a7 150 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 151 saddr = x->coaddr;
9bb182a7
YH		 152 daddr = prev_daddr;
153 }
154 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
155 saddr = prev_saddr;
66cdb3ca 156 daddr = x->coaddr;
9bb182a7 157 }
1da177e4 158
42a7b32b 159 dst = __xfrm_dst_lookup(net, tos, oif, saddr, daddr, family);
9bb182a7
YH		 160
161 if (!IS_ERR(dst)) {
162 if (prev_saddr != saddr)
163 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
164 if (prev_daddr != daddr)
165 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
166 }
1da177e4 167
66cdb3ca 168 return dst;
1da177e4 169}
1da177e4 170
1da177e4
LT		 171static inline unsigned long make_jiffies(long secs)
172{
173 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
174 return MAX_SCHEDULE_TIMEOUT-1;
175 else
a716c119 176 return secs*HZ;
1da177e4
LT		 177}
178
179static void xfrm_policy_timer(unsigned long data)
180{
3e94c2dc 181 struct xfrm_policy *xp = (struct xfrm_policy *)data;
9d729f72 182 unsigned long now = get_seconds();
1da177e4
LT		 183 long next = LONG_MAX;
184 int warn = 0;
185 int dir;
186
187 read_lock(&xp->lock);
188
ea2dea9d 189 if (unlikely(xp->walk.dead))
1da177e4
LT		 190 goto out;
191
77d8d7a6 192 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 193
194 if (xp->lft.hard_add_expires_seconds) {
195 long tmo = xp->lft.hard_add_expires_seconds +
196 xp->curlft.add_time - now;
197 if (tmo <= 0)
198 goto expired;
199 if (tmo < next)
200 next = tmo;
201 }
202 if (xp->lft.hard_use_expires_seconds) {
203 long tmo = xp->lft.hard_use_expires_seconds +
204 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
205 if (tmo <= 0)
206 goto expired;
207 if (tmo < next)
208 next = tmo;
209 }
210 if (xp->lft.soft_add_expires_seconds) {
211 long tmo = xp->lft.soft_add_expires_seconds +
212 xp->curlft.add_time - now;
213 if (tmo <= 0) {
214 warn = 1;
215 tmo = XFRM_KM_TIMEOUT;
216 }
217 if (tmo < next)
218 next = tmo;
219 }
220 if (xp->lft.soft_use_expires_seconds) {
221 long tmo = xp->lft.soft_use_expires_seconds +
222 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
223 if (tmo <= 0) {
224 warn = 1;
225 tmo = XFRM_KM_TIMEOUT;
226 }
227 if (tmo < next)
228 next = tmo;
229 }
230
231 if (warn)
6c5c8ca7 232 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 233 if (next != LONG_MAX &&
234 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
235 xfrm_pol_hold(xp);
236
237out:
238 read_unlock(&xp->lock);
239 xfrm_pol_put(xp);
240 return;
241
242expired:
243 read_unlock(&xp->lock);
4666faab 244 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 245 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 246 xfrm_pol_put(xp);
247}
248
fe1a5f03
TT		 249static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
250{
251 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
252
253 if (unlikely(pol->walk.dead))
254 flo = NULL;
255 else
256 xfrm_pol_hold(pol);
257
258 return flo;
259}
260
261static int xfrm_policy_flo_check(struct flow_cache_object *flo)
262{
263 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
264
265 return !pol->walk.dead;
266}
267
268static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
269{
270 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
271}
272
273static const struct flow_cache_ops xfrm_policy_fc_ops = {
274 .get = xfrm_policy_flo_get,
275 .check = xfrm_policy_flo_check,
276 .delete = xfrm_policy_flo_delete,
277};
1da177e4
LT		 278
279/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
280 * SPD calls.
281 */
282
0331b1f3 283struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 284{
285 struct xfrm_policy *policy;
286
0da974f4 287 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 288
289 if (policy) {
0331b1f3 290 write_pnet(&policy->xp_net, net);
12a169e7 291 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 292 INIT_HLIST_NODE(&policy->bydst);
293 INIT_HLIST_NODE(&policy->byidx);
1da177e4 294 rwlock_init(&policy->lock);
2518c7c2 295 atomic_set(&policy->refcnt, 1);
a0073fe1 296 skb_queue_head_init(&policy->polq.hold_queue);
b24b8a24
PE		 297 setup_timer(&policy->timer, xfrm_policy_timer,
298 (unsigned long)policy);
a0073fe1
SK		 299 setup_timer(&policy->polq.hold_timer, xfrm_policy_queue_process,
300 (unsigned long)policy);
fe1a5f03 301 policy->flo.ops = &xfrm_policy_fc_ops;
1da177e4
LT		 302 }
303 return policy;
304}
305EXPORT_SYMBOL(xfrm_policy_alloc);
306
56f04730
ED		 307static void xfrm_policy_destroy_rcu(struct rcu_head *head)
308{
309 struct xfrm_policy *policy = container_of(head, struct xfrm_policy, rcu);
310
311 security_xfrm_policy_free(policy->security);
312 kfree(policy);
313}
314
1da177e4
LT		 315/* Destroy xfrm_policy: descendant resources must be released to this moment. */
316
64c31b3f 317void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 318{
12a169e7 319 BUG_ON(!policy->walk.dead);
1da177e4 320
0659eea9 321 if (del_timer(&policy->timer) || del_timer(&policy->polq.hold_timer))
1da177e4
LT		 322 BUG();
323
56f04730 324 call_rcu(&policy->rcu, xfrm_policy_destroy_rcu);
1da177e4 325}
64c31b3f 326EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4 327
1365e547 328/* Rule must be locked. Release descendant resources, announce
1da177e4
LT		 329 * entry dead. The rule must be unlinked from lists to the moment.
330 */
331
332static void xfrm_policy_kill(struct xfrm_policy *policy)
333{
12a169e7 334 policy->walk.dead = 1;
1da177e4 335
285ead17 336 atomic_inc(&policy->genid);
1da177e4 337
e7d8f6cb
SK		 338 if (del_timer(&policy->polq.hold_timer))
339 xfrm_pol_put(policy);
1ee5e667 340 skb_queue_purge(&policy->polq.hold_queue);
a0073fe1 341
285ead17
TT		 342 if (del_timer(&policy->timer))
343 xfrm_pol_put(policy);
344
345 xfrm_pol_put(policy);
1da177e4
LT		 346}
347
2518c7c2
DM		 348static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
349
e92303f8 350static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 351{
e92303f8 352 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 353}
354
b58555f1
CG		 355/* calculate policy hash thresholds */
356static void __get_hash_thresh(struct net *net,
357 unsigned short family, int dir,
358 u8 *dbits, u8 *sbits)
359{
360 switch (family) {
361 case AF_INET:
362 *dbits = net->xfrm.policy_bydst[dir].dbits4;
363 *sbits = net->xfrm.policy_bydst[dir].sbits4;
364 break;
365
366 case AF_INET6:
367 *dbits = net->xfrm.policy_bydst[dir].dbits6;
368 *sbits = net->xfrm.policy_bydst[dir].sbits6;
369 break;
370
371 default:
372 *dbits = 0;
373 *sbits = 0;
374 }
375}
376
5f803b58
DM		 377static struct hlist_head *policy_hash_bysel(struct net *net,
378 const struct xfrm_selector *sel,
379 unsigned short family, int dir)
2518c7c2 380{
1121994c 381 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 382 unsigned int hash;
383 u8 dbits;
384 u8 sbits;
385
386 __get_hash_thresh(net, family, dir, &dbits, &sbits);
387 hash = __sel_hash(sel, family, hmask, dbits, sbits);
2518c7c2 388
e1e551bc
FW		 389 if (hash == hmask + 1)
390 return &net->xfrm.policy_inexact[dir];
391
392 return rcu_dereference_check(net->xfrm.policy_bydst[dir].table,
393 lockdep_is_held(&net->xfrm.xfrm_policy_lock)) + hash;
2518c7c2
DM		 394}
395
5f803b58
DM		 396static struct hlist_head *policy_hash_direct(struct net *net,
397 const xfrm_address_t *daddr,
398 const xfrm_address_t *saddr,
399 unsigned short family, int dir)
2518c7c2 400{
1121994c 401 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 402 unsigned int hash;
403 u8 dbits;
404 u8 sbits;
405
406 __get_hash_thresh(net, family, dir, &dbits, &sbits);
407 hash = __addr_hash(daddr, saddr, family, hmask, dbits, sbits);
2518c7c2 408
e1e551bc
FW		 409 return rcu_dereference_check(net->xfrm.policy_bydst[dir].table,
410 lockdep_is_held(&net->xfrm.xfrm_policy_lock)) + hash;
2518c7c2
DM		 411}
412
b58555f1
CG		 413static void xfrm_dst_hash_transfer(struct net *net,
414 struct hlist_head *list,
2518c7c2 415 struct hlist_head *ndsttable,
b58555f1
CG		 416 unsigned int nhashmask,
417 int dir)
2518c7c2 418{
b67bfe0d 419 struct hlist_node *tmp, *entry0 = NULL;
2518c7c2 420 struct xfrm_policy *pol;
b791160b 421 unsigned int h0 = 0;
b58555f1
CG		 422 u8 dbits;
423 u8 sbits;
2518c7c2 424
b791160b 425redo:
b67bfe0d 426 hlist_for_each_entry_safe(pol, tmp, list, bydst) {
2518c7c2
DM		 427 unsigned int h;
428
b58555f1 429 __get_hash_thresh(net, pol->family, dir, &dbits, &sbits);
2518c7c2 430 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
b58555f1 431 pol->family, nhashmask, dbits, sbits);
b791160b 432 if (!entry0) {
a5eefc1d
FW		 433 hlist_del_rcu(&pol->bydst);
434 hlist_add_head_rcu(&pol->bydst, ndsttable + h);
b791160b
YH		 435 h0 = h;
436 } else {
437 if (h != h0)
438 continue;
a5eefc1d
FW		 439 hlist_del_rcu(&pol->bydst);
440 hlist_add_behind_rcu(&pol->bydst, entry0);
b791160b 441 }
b67bfe0d 442 entry0 = &pol->bydst;
b791160b
YH		 443 }
444 if (!hlist_empty(list)) {
445 entry0 = NULL;
446 goto redo;
2518c7c2
DM		 447 }
448}
449
450static void xfrm_idx_hash_transfer(struct hlist_head *list,
451 struct hlist_head *nidxtable,
452 unsigned int nhashmask)
453{
b67bfe0d 454 struct hlist_node *tmp;
2518c7c2
DM		 455 struct xfrm_policy *pol;
456
b67bfe0d 457 hlist_for_each_entry_safe(pol, tmp, list, byidx) {
2518c7c2
DM		 458 unsigned int h;
459
460 h = __idx_hash(pol->index, nhashmask);
461 hlist_add_head(&pol->byidx, nidxtable+h);
462 }
463}
464
465static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
466{
467 return ((old_hmask + 1) << 1) - 1;
468}
469
66caf628 470static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 471{
66caf628 472 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 473 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
474 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
44e36b42 475 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
e1e551bc 476 struct hlist_head *odst;
2518c7c2
DM		 477 int i;
478
479 if (!ndst)
480 return;
481
9d0380df 482 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
30846090
FW		 483 write_seqcount_begin(&xfrm_policy_hash_generation);
484
485 odst = rcu_dereference_protected(net->xfrm.policy_bydst[dir].table,
486 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
2518c7c2 487
e1e551bc
FW		 488 odst = rcu_dereference_protected(net->xfrm.policy_bydst[dir].table,
489 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
490
2518c7c2 491 for (i = hmask; i >= 0; i--)
b58555f1 492 xfrm_dst_hash_transfer(net, odst + i, ndst, nhashmask, dir);
2518c7c2 493
e1e551bc 494 rcu_assign_pointer(net->xfrm.policy_bydst[dir].table, ndst);
66caf628 495 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2 496
30846090 497 write_seqcount_end(&xfrm_policy_hash_generation);
9d0380df 498 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 499
e1e551bc
FW		 500 synchronize_rcu();
501
44e36b42 502 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 503}
504
66caf628 505static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 506{
66caf628 507 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 508 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
509 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 510 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 511 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 512 int i;
513
514 if (!nidx)
515 return;
516
9d0380df 517 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 518
519 for (i = hmask; i >= 0; i--)
520 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
521
66caf628
AD		 522 net->xfrm.policy_byidx = nidx;
523 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2 524
9d0380df 525 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 526
44e36b42 527 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 528}
529
66caf628 530static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 531{
66caf628
AD		 532 unsigned int cnt = net->xfrm.policy_count[dir];
533 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 534
535 if (total)
536 *total += cnt;
537
538 if ((hmask + 1) < xfrm_policy_hashmax &&
539 cnt > hmask)
540 return 1;
541
542 return 0;
543}
544
66caf628 545static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 546{
66caf628 547 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 548
549 if ((hmask + 1) < xfrm_policy_hashmax &&
550 total > hmask)
551 return 1;
552
553 return 0;
554}
555
e071041b 556void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18 557{
e071041b
AD		 558 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
559 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
560 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
561 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
562 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
563 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
564 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18 565 si->spdhmcnt = xfrm_policy_hashmax;
ecfd6b18
JHS		 566}
567EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 568
ecfd6b18 569static DEFINE_MUTEX(hash_resize_mutex);
66caf628 570static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 571{
66caf628 572 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 573 int dir, total;
574
575 mutex_lock(&hash_resize_mutex);
576
577 total = 0;
53c2e285 578 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
66caf628
AD		 579 if (xfrm_bydst_should_resize(net, dir, &total))
580 xfrm_bydst_resize(net, dir);
2518c7c2 581 }
66caf628
AD		 582 if (xfrm_byidx_should_resize(net, total))
583 xfrm_byidx_resize(net, total);
2518c7c2
DM		 584
585 mutex_unlock(&hash_resize_mutex);
586}
587
880a6fab
CG		 588static void xfrm_hash_rebuild(struct work_struct *work)
589{
590 struct net *net = container_of(work, struct net,
591 xfrm.policy_hthresh.work);
592 unsigned int hmask;
593 struct xfrm_policy *pol;
594 struct xfrm_policy *policy;
595 struct hlist_head *chain;
596 struct hlist_head *odst;
597 struct hlist_node *newpos;
598 int i;
599 int dir;
600 unsigned seq;
601 u8 lbits4, rbits4, lbits6, rbits6;
602
603 mutex_lock(&hash_resize_mutex);
604
605 /* read selector prefixlen thresholds */
606 do {
607 seq = read_seqbegin(&net->xfrm.policy_hthresh.lock);
608
609 lbits4 = net->xfrm.policy_hthresh.lbits4;
610 rbits4 = net->xfrm.policy_hthresh.rbits4;
611 lbits6 = net->xfrm.policy_hthresh.lbits6;
612 rbits6 = net->xfrm.policy_hthresh.rbits6;
613 } while (read_seqretry(&net->xfrm.policy_hthresh.lock, seq));
614
9d0380df 615 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
880a6fab
CG		 616
617 /* reset the bydst and inexact table in all directions */
53c2e285 618 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
880a6fab
CG		 619 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
620 hmask = net->xfrm.policy_bydst[dir].hmask;
621 odst = net->xfrm.policy_bydst[dir].table;
622 for (i = hmask; i >= 0; i--)
623 INIT_HLIST_HEAD(odst + i);
624 if ((dir & XFRM_POLICY_MASK) == XFRM_POLICY_OUT) {
625 /* dir out => dst = remote, src = local */
626 net->xfrm.policy_bydst[dir].dbits4 = rbits4;
627 net->xfrm.policy_bydst[dir].sbits4 = lbits4;
628 net->xfrm.policy_bydst[dir].dbits6 = rbits6;
629 net->xfrm.policy_bydst[dir].sbits6 = lbits6;
630 } else {
631 /* dir in/fwd => dst = local, src = remote */
632 net->xfrm.policy_bydst[dir].dbits4 = lbits4;
633 net->xfrm.policy_bydst[dir].sbits4 = rbits4;
634 net->xfrm.policy_bydst[dir].dbits6 = lbits6;
635 net->xfrm.policy_bydst[dir].sbits6 = rbits6;
636 }
637 }
638
639 /* re-insert all policies by order of creation */
640 list_for_each_entry_reverse(policy, &net->xfrm.policy_all, walk.all) {
6916fb3b
TB		 641 if (xfrm_policy_id2dir(policy->index) >= XFRM_POLICY_MAX) {
642 /* skip socket policies */
643 continue;
644 }
880a6fab
CG		 645 newpos = NULL;
646 chain = policy_hash_bysel(net, &policy->selector,
647 policy->family,
648 xfrm_policy_id2dir(policy->index));
649 hlist_for_each_entry(pol, chain, bydst) {
650 if (policy->priority >= pol->priority)
651 newpos = &pol->bydst;
652 else
653 break;
654 }
655 if (newpos)
656 hlist_add_behind(&policy->bydst, newpos);
657 else
658 hlist_add_head(&policy->bydst, chain);
659 }
660
9d0380df 661 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
880a6fab
CG		 662
663 mutex_unlock(&hash_resize_mutex);
664}
665
666void xfrm_policy_hash_rebuild(struct net *net)
667{
668 schedule_work(&net->xfrm.policy_hthresh.work);
669}
670EXPORT_SYMBOL(xfrm_policy_hash_rebuild);
671
1da177e4
LT		 672/* Generate new index... KAME seems to generate them ordered by cost
673 * of an absolute inpredictability of ordering of rules. This will not pass. */
e682adf0 674static u32 xfrm_gen_index(struct net *net, int dir, u32 index)
1da177e4 675{
1da177e4
LT		 676 static u32 idx_generator;
677
678 for (;;) {
2518c7c2
DM		 679 struct hlist_head *list;
680 struct xfrm_policy *p;
681 u32 idx;
682 int found;
683
e682adf0
FD		 684 if (!index) {
685 idx = (idx_generator | dir);
686 idx_generator += 8;
687 } else {
688 idx = index;
689 index = 0;
690 }
691
1da177e4
LT		 692 if (idx == 0)
693 idx = 8;
1121994c 694 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 695 found = 0;
b67bfe0d 696 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 697 if (p->index == idx) {
698 found = 1;
1da177e4 699 break;
2518c7c2 700 }
1da177e4 701 }
2518c7c2 702 if (!found)
1da177e4
LT		 703 return idx;
704 }
705}
706
2518c7c2
DM		 707static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
708{
709 u32 *p1 = (u32 *) s1;
710 u32 *p2 = (u32 *) s2;
711 int len = sizeof(struct xfrm_selector) / sizeof(u32);
712 int i;
713
714 for (i = 0; i < len; i++) {
715 if (p1[i] != p2[i])
716 return 1;
717 }
718
719 return 0;
720}
721
a0073fe1
SK		 722static void xfrm_policy_requeue(struct xfrm_policy *old,
723 struct xfrm_policy *new)
724{
725 struct xfrm_policy_queue *pq = &old->polq;
726 struct sk_buff_head list;
727
de2ad486
LR		 728 if (skb_queue_empty(&pq->hold_queue))
729 return;
730
a0073fe1
SK		 731 __skb_queue_head_init(&list);
732
733 spin_lock_bh(&pq->hold_queue.lock);
734 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 735 if (del_timer(&pq->hold_timer))
736 xfrm_pol_put(old);
a0073fe1
SK		 737 spin_unlock_bh(&pq->hold_queue.lock);
738
a0073fe1
SK		 739 pq = &new->polq;
740
741 spin_lock_bh(&pq->hold_queue.lock);
742 skb_queue_splice(&list, &pq->hold_queue);
743 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 744 if (!mod_timer(&pq->hold_timer, jiffies))
745 xfrm_pol_hold(new);
a0073fe1
SK		 746 spin_unlock_bh(&pq->hold_queue.lock);
747}
748
7cb8a939
SK		 749static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
750 struct xfrm_policy *pol)
751{
752 u32 mark = policy->mark.v & policy->mark.m;
753
754 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
755 return true;
756
757 if ((mark & pol->mark.m) == pol->mark.v &&
758 policy->priority == pol->priority)
759 return true;
760
761 return false;
762}
763
1da177e4
LT		 764int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
765{
1121994c 766 struct net *net = xp_net(policy);
2518c7c2
DM		 767 struct xfrm_policy *pol;
768 struct xfrm_policy *delpol;
769 struct hlist_head *chain;
b67bfe0d 770 struct hlist_node *newpos;
1da177e4 771
9d0380df 772 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
1121994c 773 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 774 delpol = NULL;
775 newpos = NULL;
b67bfe0d 776 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 777 if (pol->type == policy->type &&
2518c7c2 778 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 779 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 780 xfrm_sec_ctx_match(pol->security, policy->security) &&
781 !WARN_ON(delpol)) {
1da177e4 782 if (excl) {
9d0380df 783 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 784 return -EEXIST;
785 }
1da177e4
LT		 786 delpol = pol;
787 if (policy->priority > pol->priority)
788 continue;
789 } else if (policy->priority >= pol->priority) {
a6c7ab55 790 newpos = &pol->bydst;
1da177e4
LT		 791 continue;
792 }
1da177e4
LT		 793 if (delpol)
794 break;
1da177e4
LT		 795 }
796 if (newpos)
1d023284 797 hlist_add_behind(&policy->bydst, newpos);
2518c7c2
DM		 798 else
799 hlist_add_head(&policy->bydst, chain);
12bfa8bd 800 __xfrm_policy_link(policy, dir);
ca925cf1 801 atomic_inc(&net->xfrm.flow_cache_genid);
ca4c3fc2 802
803 /* After previous checking, family can either be AF_INET or AF_INET6 */
804 if (policy->family == AF_INET)
805 rt_genid_bump_ipv4(net);
806 else
807 rt_genid_bump_ipv6(net);
808
a0073fe1
SK		 809 if (delpol) {
810 xfrm_policy_requeue(delpol, policy);
29fa0b30 811 __xfrm_policy_unlink(delpol, dir);
a0073fe1 812 }
e682adf0 813 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index);
1121994c 814 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 815 policy->curlft.add_time = get_seconds();
1da177e4
LT		 816 policy->curlft.use_time = 0;
817 if (!mod_timer(&policy->timer, jiffies + HZ))
818 xfrm_pol_hold(policy);
9d0380df 819 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 820
9b78a82c 821 if (delpol)
1da177e4 822 xfrm_policy_kill(delpol);
1121994c
AD		 823 else if (xfrm_bydst_should_resize(net, dir, NULL))
824 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 825
1da177e4
LT		 826 return 0;
827}
828EXPORT_SYMBOL(xfrm_policy_insert);
829
8ca2e93b
JHS		 830struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
831 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 832 struct xfrm_sec_ctx *ctx, int delete,
833 int *err)
1da177e4 834{
2518c7c2
DM		 835 struct xfrm_policy *pol, *ret;
836 struct hlist_head *chain;
1da177e4 837
ef41aaa0 838 *err = 0;
9d0380df 839 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 840 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 841 ret = NULL;
b67bfe0d 842 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 843 if (pol->type == type &&
34f8d884 844 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 845 !selector_cmp(sel, &pol->selector) &&
846 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 847 xfrm_pol_hold(pol);
2518c7c2 848 if (delete) {
03e1ad7b
PM		 849 *err = security_xfrm_policy_delete(
850 pol->security);
ef41aaa0 851 if (*err) {
9d0380df 852 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 853 return pol;
854 }
29fa0b30 855 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 856 }
857 ret = pol;
1da177e4
LT		 858 break;
859 }
860 }
9d0380df 861 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 862
fe1a5f03 863 if (ret && delete)
2518c7c2 864 xfrm_policy_kill(ret);
2518c7c2 865 return ret;
1da177e4 866}
df71837d 867EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 868
8ca2e93b
JHS		 869struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
870 int dir, u32 id, int delete, int *err)
1da177e4 871{
2518c7c2
DM		 872 struct xfrm_policy *pol, *ret;
873 struct hlist_head *chain;
1da177e4 874
b5505c6e
HX		 875 *err = -ENOENT;
876 if (xfrm_policy_id2dir(id) != dir)
877 return NULL;
878
ef41aaa0 879 *err = 0;
9d0380df 880 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 881 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 882 ret = NULL;
b67bfe0d 883 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 884 if (pol->type == type && pol->index == id &&
885 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 886 xfrm_pol_hold(pol);
2518c7c2 887 if (delete) {
03e1ad7b
PM		 888 *err = security_xfrm_policy_delete(
889 pol->security);
ef41aaa0 890 if (*err) {
9d0380df 891 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 892 return pol;
893 }
29fa0b30 894 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 895 }
896 ret = pol;
1da177e4
LT		 897 break;
898 }
899 }
9d0380df 900 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 901
fe1a5f03 902 if (ret && delete)
2518c7c2 903 xfrm_policy_kill(ret);
2518c7c2 904 return ret;
1da177e4
LT		 905}
906EXPORT_SYMBOL(xfrm_policy_byid);
907
4aa2e62c
JL		 908#ifdef CONFIG_SECURITY_NETWORK_XFRM
909static inline int
2e71029e 910xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
1da177e4 911{
4aa2e62c
JL		 912 int dir, err = 0;
913
914 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
915 struct xfrm_policy *pol;
4aa2e62c
JL		 916 int i;
917
b67bfe0d 918 hlist_for_each_entry(pol,
33ffbbd5 919 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 920 if (pol->type != type)
921 continue;
03e1ad7b 922 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 923 if (err) {
2e71029e 924 xfrm_audit_policy_delete(pol, 0, task_valid);
4aa2e62c
JL		 925 return err;
926 }
7dc12d6d 927 }
33ffbbd5 928 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 929 hlist_for_each_entry(pol,
33ffbbd5 930 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 931 bydst) {
932 if (pol->type != type)
933 continue;
03e1ad7b
PM		 934 err = security_xfrm_policy_delete(
935 pol->security);
4aa2e62c 936 if (err) {
ab5f5e8b 937 xfrm_audit_policy_delete(pol, 0,
2e71029e 938 task_valid);
4aa2e62c
JL		 939 return err;
940 }
941 }
942 }
943 }
944 return err;
945}
946#else
947static inline int
2e71029e 948xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
4aa2e62c
JL		 949{
950 return 0;
951}
952#endif
953
2e71029e 954int xfrm_policy_flush(struct net *net, u8 type, bool task_valid)
4aa2e62c 955{
2f1eb65f 956 int dir, err = 0, cnt = 0;
1da177e4 957
9d0380df 958 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 959
2e71029e 960 err = xfrm_policy_flush_secctx_check(net, type, task_valid);
4aa2e62c
JL		 961 if (err)
962 goto out;
963
1da177e4 964 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 965 struct xfrm_policy *pol;
29fa0b30 966 int i;
2518c7c2
DM		 967
968 again1:
b67bfe0d 969 hlist_for_each_entry(pol,
33ffbbd5 970 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 971 if (pol->type != type)
972 continue;
ea2dea9d 973 __xfrm_policy_unlink(pol, dir);
9d0380df 974 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 975 cnt++;
1da177e4 976
2e71029e 977 xfrm_audit_policy_delete(pol, 1, task_valid);
161a09e7 978
2518c7c2 979 xfrm_policy_kill(pol);
1da177e4 980
4141b36a 981 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 982 goto again1;
983 }
984
33ffbbd5 985 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 986 again2:
b67bfe0d 987 hlist_for_each_entry(pol,
33ffbbd5 988 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 989 bydst) {
990 if (pol->type != type)
991 continue;
ea2dea9d 992 __xfrm_policy_unlink(pol, dir);
9d0380df 993 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 994 cnt++;
2518c7c2 995
2e71029e 996 xfrm_audit_policy_delete(pol, 1, task_valid);
2518c7c2
DM		 997 xfrm_policy_kill(pol);
998
9d0380df 999 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 1000 goto again2;
1001 }
1da177e4 1002 }
2518c7c2 1003
1da177e4 1004 }
2f1eb65f
JHS		 1005 if (!cnt)
1006 err = -ESRCH;
4aa2e62c 1007out:
9d0380df 1008 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
35db0691
XL		 1009
1010 if (cnt)
1011 xfrm_garbage_collect(net);
1012
4aa2e62c 1013 return err;
1da177e4
LT		 1014}
1015EXPORT_SYMBOL(xfrm_policy_flush);
1016
cdcbca7c 1017int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 1018 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 1019 void *data)
1020{
12a169e7
HX		 1021 struct xfrm_policy *pol;
1022 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 1023 int error = 0;
1024
1025 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
1026 walk->type != XFRM_POLICY_TYPE_ANY)
1027 return -EINVAL;
1da177e4 1028
12a169e7 1029 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 1030 return 0;
1031
9d0380df 1032 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7 1033 if (list_empty(&walk->walk.all))
cdcbca7c 1034 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7 1035 else
80077702
LR		 1036 x = list_first_entry(&walk->walk.all,
1037 struct xfrm_policy_walk_entry, all);
1038
cdcbca7c 1039 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 1040 if (x->dead)
4c563f76 1041 continue;
12a169e7
HX		 1042 pol = container_of(x, struct xfrm_policy, walk);
1043 if (walk->type != XFRM_POLICY_TYPE_ANY &&
1044 walk->type != pol->type)
1045 continue;
1046 error = func(pol, xfrm_policy_id2dir(pol->index),
1047 walk->seq, data);
1048 if (error) {
1049 list_move_tail(&walk->walk.all, &x->all);
1050 goto out;
2518c7c2 1051 }
12a169e7 1052 walk->seq++;
1da177e4 1053 }
12a169e7 1054 if (walk->seq == 0) {
baf5d743
JHS		 1055 error = -ENOENT;
1056 goto out;
1057 }
12a169e7 1058 list_del_init(&walk->walk.all);
1da177e4 1059out:
9d0380df 1060 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1061 return error;
1062}
1063EXPORT_SYMBOL(xfrm_policy_walk);
1064
12a169e7
HX		 1065void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
1066{
1067 INIT_LIST_HEAD(&walk->walk.all);
1068 walk->walk.dead = 1;
1069 walk->type = type;
1070 walk->seq = 0;
1071}
1072EXPORT_SYMBOL(xfrm_policy_walk_init);
1073
283bc9f3 1074void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net)
12a169e7
HX		 1075{
1076 if (list_empty(&walk->walk.all))
1077 return;
1078
9d0380df 1079 spin_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */
12a169e7 1080 list_del(&walk->walk.all);
9d0380df 1081 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7
HX		 1082}
1083EXPORT_SYMBOL(xfrm_policy_walk_done);
1084
134b0fc5
JM		 1085/*
1086 * Find policy to apply to this flow.
1087 *
1088 * Returns 0 if policy found, else an -errno.
1089 */
f299d557
DM		 1090static int xfrm_policy_match(const struct xfrm_policy *pol,
1091 const struct flowi *fl,
2518c7c2 1092 u8 type, u16 family, int dir)
1da177e4 1093{
f299d557 1094 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 1095 int ret = -ESRCH;
1096 bool match;
1da177e4 1097
2518c7c2 1098 if (pol->family != family ||
1d28f42c 1099 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 1100 pol->type != type)
134b0fc5 1101 return ret;
1da177e4 1102
2518c7c2 1103 match = xfrm_selector_match(sel, fl, family);
134b0fc5 1104 if (match)
1d28f42c 1105 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 1106 dir);
2518c7c2 1107
134b0fc5 1108 return ret;
2518c7c2 1109}
1da177e4 1110
52479b62 1111static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 1112 const struct flowi *fl,
2518c7c2
DM		 1113 u16 family, u8 dir)
1114{
134b0fc5 1115 int err;
2518c7c2 1116 struct xfrm_policy *pol, *ret;
0b597e7e 1117 const xfrm_address_t *daddr, *saddr;
2518c7c2 1118 struct hlist_head *chain;
30846090
FW		 1119 unsigned int sequence;
1120 u32 priority;
df71837d 1121
2518c7c2
DM		 1122 daddr = xfrm_flowi_daddr(fl, family);
1123 saddr = xfrm_flowi_saddr(fl, family);
1124 if (unlikely(!daddr || !saddr))
1125 return NULL;
1126
a7c44247 1127 rcu_read_lock();
30846090
FW		 1128 retry:
1129 do {
1130 sequence = read_seqcount_begin(&xfrm_policy_hash_generation);
1131 chain = policy_hash_direct(net, daddr, saddr, family, dir);
1132 } while (read_seqcount_retry(&xfrm_policy_hash_generation, sequence));
1133
1134 priority = ~0U;
2518c7c2 1135 ret = NULL;
a5eefc1d 1136 hlist_for_each_entry_rcu(pol, chain, bydst) {
134b0fc5
JM		 1137 err = xfrm_policy_match(pol, fl, type, family, dir);
1138 if (err) {
1139 if (err == -ESRCH)
1140 continue;
1141 else {
1142 ret = ERR_PTR(err);
1143 goto fail;
1144 }
1145 } else {
2518c7c2 1146 ret = pol;
acba48e1 1147 priority = ret->priority;
2518c7c2
DM		 1148 break;
1149 }
1150 }
52479b62 1151 chain = &net->xfrm.policy_inexact[dir];
a5eefc1d 1152 hlist_for_each_entry_rcu(pol, chain, bydst) {
8faf491e
LR		 1153 if ((pol->priority >= priority) && ret)
1154 break;
1155
134b0fc5
JM		 1156 err = xfrm_policy_match(pol, fl, type, family, dir);
1157 if (err) {
1158 if (err == -ESRCH)
1159 continue;
1160 else {
1161 ret = ERR_PTR(err);
1162 goto fail;
1163 }
8faf491e 1164 } else {
acba48e1
DM		 1165 ret = pol;
1166 break;
1da177e4
LT		 1167 }
1168 }
586f2eb4 1169
30846090
FW		 1170 if (read_seqcount_retry(&xfrm_policy_hash_generation, sequence))
1171 goto retry;
1172
e37cc8ad
FW		 1173 if (ret && !xfrm_pol_hold_rcu(ret))
1174 goto retry;
134b0fc5 1175fail:
a7c44247 1176 rcu_read_unlock();
4e81bb83 1177
2518c7c2 1178 return ret;
4e81bb83
MN		 1179}
1180
80c802f3 1181static struct xfrm_policy *
73ff93cd 1182__xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1183{
1184#ifdef CONFIG_XFRM_SUB_POLICY
1185 struct xfrm_policy *pol;
1186
1187 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1188 if (pol != NULL)
1189 return pol;
1190#endif
1191 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1192}
1193
b5fb82c4
BZ		 1194static int flow_to_policy_dir(int dir)
1195{
1196 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
1197 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1198 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1199 return dir;
1200
1201 switch (dir) {
1202 default:
1203 case FLOW_DIR_IN:
1204 return XFRM_POLICY_IN;
1205 case FLOW_DIR_OUT:
1206 return XFRM_POLICY_OUT;
1207 case FLOW_DIR_FWD:
1208 return XFRM_POLICY_FWD;
1209 }
1210}
1211
fe1a5f03 1212static struct flow_cache_object *
dee9f4bc 1213xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
fe1a5f03 1214 u8 dir, struct flow_cache_object *old_obj, void *ctx)
4e81bb83
MN		 1215{
1216 struct xfrm_policy *pol;
fe1a5f03
TT		 1217
1218 if (old_obj)
1219 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
4e81bb83 1220
b5fb82c4 1221 pol = __xfrm_policy_lookup(net, fl, family, flow_to_policy_dir(dir));
80c802f3 1222 if (IS_ERR_OR_NULL(pol))
fe1a5f03 1223 return ERR_CAST(pol);
fe1a5f03 1224
fe1a5f03
TT		 1225 /* Resolver returns two references:
1226 * one for cache and one for caller of flow_cache_lookup() */
1227 xfrm_pol_hold(pol);
1228
1229 return &pol->flo;
1da177e4
LT		 1230}
1231
df71837d
TJ		 1232static inline int policy_to_flow_dir(int dir)
1233{
1234 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1235 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1236 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1237 return dir;
1238 switch (dir) {
1239 default:
1240 case XFRM_POLICY_IN:
1241 return FLOW_DIR_IN;
1242 case XFRM_POLICY_OUT:
1243 return FLOW_DIR_OUT;
1244 case XFRM_POLICY_FWD:
1245 return FLOW_DIR_FWD;
3ff50b79 1246 }
df71837d
TJ		 1247}
1248
6f9c9615 1249static struct xfrm_policy *xfrm_sk_policy_lookup(const struct sock *sk, int dir,
4c86d777 1250 const struct flowi *fl, u16 family)
1da177e4
LT		 1251{
1252 struct xfrm_policy *pol;
1253
d188ba86 1254 rcu_read_lock();
ae33786f 1255 again:
d188ba86
ED		 1256 pol = rcu_dereference(sk->sk_policy[dir]);
1257 if (pol != NULL) {
4c86d777 1258 bool match = xfrm_selector_match(&pol->selector, fl, family);
a716c119 1259 int err = 0;
df71837d 1260
3bccfbc7 1261 if (match) {
34f8d884
JHS		 1262 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1263 pol = NULL;
1264 goto out;
1265 }
03e1ad7b 1266 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1267 fl->flowi_secid,
03e1ad7b 1268 policy_to_flow_dir(dir));
330e832a
FW		 1269 if (!err) {
1270 if (!xfrm_pol_hold_rcu(pol))
1271 goto again;
1272 } else if (err == -ESRCH) {
3bccfbc7 1273 pol = NULL;
330e832a 1274 } else {
3bccfbc7 1275 pol = ERR_PTR(err);
330e832a 1276 }
3bccfbc7 1277 } else
1da177e4
LT		 1278 pol = NULL;
1279 }
34f8d884 1280out:
d188ba86 1281 rcu_read_unlock();
1da177e4
LT		 1282 return pol;
1283}
1284
1285static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1286{
98806f75 1287 struct net *net = xp_net(pol);
4e81bb83 1288
98806f75 1289 list_add(&pol->walk.all, &net->xfrm.policy_all);
98806f75 1290 net->xfrm.policy_count[dir]++;
1da177e4
LT		 1291 xfrm_pol_hold(pol);
1292}
1293
1294static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1295 int dir)
1296{
98806f75
AD		 1297 struct net *net = xp_net(pol);
1298
53c2e285 1299 if (list_empty(&pol->walk.all))
2518c7c2 1300 return NULL;
1da177e4 1301
53c2e285
HX		 1302 /* Socket policies are not hashed. */
1303 if (!hlist_unhashed(&pol->bydst)) {
a5eefc1d 1304 hlist_del_rcu(&pol->bydst);
53c2e285
HX		 1305 hlist_del(&pol->byidx);
1306 }
1307
1308 list_del_init(&pol->walk.all);
98806f75 1309 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1310
1311 return pol;
1da177e4
LT		 1312}
1313
53c2e285
HX		 1314static void xfrm_sk_policy_link(struct xfrm_policy *pol, int dir)
1315{
1316 __xfrm_policy_link(pol, XFRM_POLICY_MAX + dir);
1317}
1318
1319static void xfrm_sk_policy_unlink(struct xfrm_policy *pol, int dir)
1320{
1321 __xfrm_policy_unlink(pol, XFRM_POLICY_MAX + dir);
1322}
1323
4666faab 1324int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4 1325{
283bc9f3
FD		 1326 struct net *net = xp_net(pol);
1327
9d0380df 1328 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1329 pol = __xfrm_policy_unlink(pol, dir);
9d0380df 1330 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1331 if (pol) {
1da177e4 1332 xfrm_policy_kill(pol);
4666faab 1333 return 0;
1da177e4 1334 }
4666faab 1335 return -ENOENT;
1da177e4 1336}
a70fcb0b 1337EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1338
1339int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1340{
1121994c 1341 struct net *net = xp_net(pol);
1da177e4
LT		 1342 struct xfrm_policy *old_pol;
1343
4e81bb83
MN		 1344#ifdef CONFIG_XFRM_SUB_POLICY
1345 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1346 return -EINVAL;
1347#endif
1348
9d0380df 1349 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
d188ba86
ED		 1350 old_pol = rcu_dereference_protected(sk->sk_policy[dir],
1351 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
1da177e4 1352 if (pol) {
9d729f72 1353 pol->curlft.add_time = get_seconds();
e682adf0 1354 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
53c2e285 1355 xfrm_sk_policy_link(pol, dir);
1da177e4 1356 }
d188ba86 1357 rcu_assign_pointer(sk->sk_policy[dir], pol);
a0073fe1
SK		 1358 if (old_pol) {
1359 if (pol)
1360 xfrm_policy_requeue(old_pol, pol);
1361
ea2dea9d
TT		 1362 /* Unlinking succeeds always. This is the only function
1363 * allowed to delete or replace socket policy.
1364 */
53c2e285 1365 xfrm_sk_policy_unlink(old_pol, dir);
a0073fe1 1366 }
9d0380df 1367 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1368
1369 if (old_pol) {
1370 xfrm_policy_kill(old_pol);
1371 }
1372 return 0;
1373}
1374
d3e40a9f 1375static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1376{
0331b1f3 1377 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
283bc9f3 1378 struct net *net = xp_net(old);
1da177e4
LT		 1379
1380 if (newp) {
1381 newp->selector = old->selector;
03e1ad7b
PM		 1382 if (security_xfrm_policy_clone(old->security,
1383 &newp->security)) {
df71837d
TJ		 1384 kfree(newp);
1385 return NULL; /* ENOMEM */
1386 }
1da177e4
LT		 1387 newp->lft = old->lft;
1388 newp->curlft = old->curlft;
fb977e2c 1389 newp->mark = old->mark;
1da177e4
LT		 1390 newp->action = old->action;
1391 newp->flags = old->flags;
1392 newp->xfrm_nr = old->xfrm_nr;
1393 newp->index = old->index;
4e81bb83 1394 newp->type = old->type;
1da177e4
LT		 1395 memcpy(newp->xfrm_vec, old->xfrm_vec,
1396 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
9d0380df 1397 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
53c2e285 1398 xfrm_sk_policy_link(newp, dir);
9d0380df 1399 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1400 xfrm_pol_put(newp);
1401 }
1402 return newp;
1403}
1404
d188ba86 1405int __xfrm_sk_clone_policy(struct sock *sk, const struct sock *osk)
1da177e4 1406{
d188ba86
ED		 1407 const struct xfrm_policy *p;
1408 struct xfrm_policy *np;
1409 int i, ret = 0;
1410
1411 rcu_read_lock();
1412 for (i = 0; i < 2; i++) {
1413 p = rcu_dereference(osk->sk_policy[i]);
1414 if (p) {
1415 np = clone_policy(p, i);
1416 if (unlikely(!np)) {
1417 ret = -ENOMEM;
1418 break;
1419 }
1420 rcu_assign_pointer(sk->sk_policy[i], np);
1421 }
1422 }
1423 rcu_read_unlock();
1424 return ret;
1da177e4
LT		 1425}
1426
a1e59abf 1427static int
42a7b32b
DA		 1428xfrm_get_saddr(struct net *net, int oif, xfrm_address_t *local,
1429 xfrm_address_t *remote, unsigned short family)
a1e59abf
PM		 1430{
1431 int err;
37b10383 1432 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
a1e59abf
PM		 1433
1434 if (unlikely(afinfo == NULL))
1435 return -EINVAL;
42a7b32b 1436 err = afinfo->get_saddr(net, oif, local, remote);
bdba9fe0 1437 rcu_read_unlock();
a1e59abf
PM		 1438 return err;
1439}
1440
1da177e4
LT		 1441/* Resolve list of templates for the flow, given policy. */
1442
1443static int
a6c2e611
DM		 1444xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1445 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1446{
fbda33b2 1447 struct net *net = xp_net(policy);
1da177e4
LT		 1448 int nx;
1449 int i, error;
1450 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1451 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1452 xfrm_address_t tmp;
1da177e4 1453
9b7a787d 1454 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) {
1da177e4
LT		 1455 struct xfrm_state *x;
1456 xfrm_address_t *remote = daddr;
1457 xfrm_address_t *local = saddr;
1458 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1459
48b8d783
JK		 1460 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1461 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1462 remote = &tmpl->id.daddr;
1463 local = &tmpl->saddr;
8444cf71 1464 if (xfrm_addr_any(local, tmpl->encap_family)) {
42a7b32b
DA		 1465 error = xfrm_get_saddr(net, fl->flowi_oif,
1466 &tmp, remote,
1467 tmpl->encap_family);
a1e59abf
PM		 1468 if (error)
1469 goto fail;
1470 local = &tmp;
1471 }
1da177e4
LT		 1472 }
1473
1474 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1475
1476 if (x && x->km.state == XFRM_STATE_VALID) {
1477 xfrm[nx++] = x;
1478 daddr = remote;
1479 saddr = local;
1480 continue;
1481 }
1482 if (x) {
1483 error = (x->km.state == XFRM_STATE_ERROR ?
1484 -EINVAL : -EAGAIN);
1485 xfrm_state_put(x);
42054569 1486 } else if (error == -ESRCH) {
a4322266 1487 error = -EAGAIN;
42054569 1488 }
1da177e4
LT		 1489
1490 if (!tmpl->optional)
1491 goto fail;
1492 }
1493 return nx;
1494
1495fail:
9b7a787d 1496 for (nx--; nx >= 0; nx--)
1da177e4
LT		 1497 xfrm_state_put(xfrm[nx]);
1498 return error;
1499}
1500
4e81bb83 1501static int
a6c2e611
DM		 1502xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1503 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1504{
41a49cc3
MN		 1505 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1506 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1507 int cnx = 0;
1508 int error;
1509 int ret;
1510 int i;
1511
1512 for (i = 0; i < npols; i++) {
1513 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1514 error = -ENOBUFS;
1515 goto fail;
1516 }
41a49cc3
MN		 1517
1518 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1519 if (ret < 0) {
1520 error = ret;
1521 goto fail;
1522 } else
1523 cnx += ret;
1524 }
1525
41a49cc3
MN		 1526 /* found states are sorted for outbound processing */
1527 if (npols > 1)
1528 xfrm_state_sort(xfrm, tpp, cnx, family);
1529
4e81bb83
MN		 1530 return cnx;
1531
1532 fail:
9b7a787d 1533 for (cnx--; cnx >= 0; cnx--)
41a49cc3 1534 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1535 return error;
1536
1537}
1538
f5e2bb4f 1539static int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286 1540{
37b10383 1541 const struct xfrm_policy_afinfo *afinfo;
f5e2bb4f 1542 int tos = 0;
1da177e4 1543
f5e2bb4f
FW		 1544 afinfo = xfrm_policy_get_afinfo(family);
1545 tos = afinfo ? afinfo->get_tos(fl) : 0;
25ee3286 1546
bdba9fe0 1547 rcu_read_unlock();
25ee3286
HX		 1548
1549 return tos;
1550}
1551
80c802f3
TT		 1552static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1553{
1554 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1555 struct dst_entry *dst = &xdst->u.dst;
1556
1557 if (xdst->route == NULL) {
1558 /* Dummy bundle - if it has xfrms we were not
1559 * able to build bundle as template resolution failed.
1560 * It means we need to try again resolving. */
1561 if (xdst->num_xfrms > 0)
1562 return NULL;
a0073fe1
SK		 1563 } else if (dst->flags & DST_XFRM_QUEUE) {
1564 return NULL;
80c802f3
TT		 1565 } else {
1566 /* Real bundle */
1567 if (stale_bundle(dst))
1568 return NULL;
1569 }
1570
1571 dst_hold(dst);
1572 return flo;
1573}
1574
1575static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1576{
1577 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1578 struct dst_entry *dst = &xdst->u.dst;
1579
1580 if (!xdst->route)
1581 return 0;
1582 if (stale_bundle(dst))
1583 return 0;
1584
1585 return 1;
1586}
1587
1588static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1589{
1590 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1591 struct dst_entry *dst = &xdst->u.dst;
1592
52df157f
WW		 1593 /* Mark DST_OBSOLETE_DEAD to fail the next xfrm_dst_check() */
1594 dst->obsolete = DST_OBSOLETE_DEAD;
1595 dst_release_immediate(dst);
80c802f3
TT		 1596}
1597
1598static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1599 .get = xfrm_bundle_flo_get,
1600 .check = xfrm_bundle_flo_check,
1601 .delete = xfrm_bundle_flo_delete,
1602};
1603
d7c7544c 1604static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1605{
37b10383 1606 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1607 struct dst_ops *dst_ops;
25ee3286
HX		 1608 struct xfrm_dst *xdst;
1609
1610 if (!afinfo)
1611 return ERR_PTR(-EINVAL);
1612
d7c7544c
AD		 1613 switch (family) {
1614 case AF_INET:
1615 dst_ops = &net->xfrm.xfrm4_dst_ops;
1616 break;
dfd56b8b 1617#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1618 case AF_INET6:
1619 dst_ops = &net->xfrm.xfrm6_dst_ops;
1620 break;
1621#endif
1622 default:
1623 BUG();
1624 }
52df157f 1625 xdst = dst_alloc(dst_ops, NULL, 1, DST_OBSOLETE_NONE, DST_NOGC);
25ee3286 1626
d4cae562 1627 if (likely(xdst)) {
141e369d
SK		 1628 struct dst_entry *dst = &xdst->u.dst;
1629
1630 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
0b150932 1631 xdst->flo.ops = &xfrm_bundle_fc_ops;
d4cae562 1632 } else
0b150932 1633 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1634
bdba9fe0 1635 rcu_read_unlock();
d4cae562 1636
25ee3286
HX		 1637 return xdst;
1638}
1639
a1b05140
MN		 1640static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1641 int nfheader_len)
1642{
37b10383 1643 const struct xfrm_policy_afinfo *afinfo =
a1b05140
MN		 1644 xfrm_policy_get_afinfo(dst->ops->family);
1645 int err;
1646
1647 if (!afinfo)
1648 return -EINVAL;
1649
1650 err = afinfo->init_path(path, dst, nfheader_len);
1651
bdba9fe0 1652 rcu_read_unlock();
a1b05140
MN		 1653
1654 return err;
1655}
1656
87c1e12b 1657static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1658 const struct flowi *fl)
25ee3286 1659{
37b10383 1660 const struct xfrm_policy_afinfo *afinfo =
25ee3286
HX		 1661 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1662 int err;
1663
1664 if (!afinfo)
1da177e4 1665 return -EINVAL;
25ee3286 1666
87c1e12b 1667 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1668
bdba9fe0 1669 rcu_read_unlock();
25ee3286 1670
1da177e4
LT		 1671 return err;
1672}
1673
80c802f3 1674
25ee3286
HX		 1675/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1676 * all the metrics... Shortly, bundle a bundle.
1677 */
1678
1679static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1680 struct xfrm_state **xfrm, int nx,
98313ada 1681 const struct flowi *fl,
25ee3286
HX		 1682 struct dst_entry *dst)
1683{
d7c7544c 1684 struct net *net = xp_net(policy);
25ee3286
HX		 1685 unsigned long now = jiffies;
1686 struct net_device *dev;
43a4dea4 1687 struct xfrm_mode *inner_mode;
25ee3286
HX		 1688 struct dst_entry *dst_prev = NULL;
1689 struct dst_entry *dst0 = NULL;
1690 int i = 0;
1691 int err;
1692 int header_len = 0;
a1b05140 1693 int nfheader_len = 0;
25ee3286
HX		 1694 int trailer_len = 0;
1695 int tos;
1696 int family = policy->selector.family;
9bb182a7
YH		 1697 xfrm_address_t saddr, daddr;
1698
1699 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1700
1701 tos = xfrm_get_tos(fl, family);
25ee3286
HX		 1702
1703 dst_hold(dst);
1704
1705 for (; i < nx; i++) {
d7c7544c 1706 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1707 struct dst_entry *dst1 = &xdst->u.dst;
1708
1709 err = PTR_ERR(xdst);
1710 if (IS_ERR(xdst)) {
1711 dst_release(dst);
1712 goto put_states;
1713 }
1714
43a4dea4
SK		 1715 if (xfrm[i]->sel.family == AF_UNSPEC) {
1716 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1717 xfrm_af2proto(family));
1718 if (!inner_mode) {
1719 err = -EAFNOSUPPORT;
1720 dst_release(dst);
1721 goto put_states;
1722 }
1723 } else
1724 inner_mode = xfrm[i]->inner_mode;
1725
25ee3286
HX		 1726 if (!dst_prev)
1727 dst0 = dst1;
52df157f
WW		 1728 else
1729 /* Ref count is taken during xfrm_alloc_dst()
1730 * No need to do dst_clone() on dst1
1731 */
1732 dst_prev->child = dst1;
25ee3286
HX		 1733
1734 xdst->route = dst;
defb3519 1735 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1736
1737 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1738 family = xfrm[i]->props.family;
42a7b32b
DA		 1739 dst = xfrm_dst_lookup(xfrm[i], tos, fl->flowi_oif,
1740 &saddr, &daddr, family);
25ee3286
HX		 1741 err = PTR_ERR(dst);
1742 if (IS_ERR(dst))
1743 goto put_states;
1744 } else
1745 dst_hold(dst);
1746
1747 dst1->xfrm = xfrm[i];
80c802f3 1748 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1749
f5b0a874 1750 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1751 dst1->flags |= DST_HOST;
1752 dst1->lastuse = now;
1753
1754 dst1->input = dst_discard;
43a4dea4 1755 dst1->output = inner_mode->afinfo->output;
25ee3286
HX		 1756
1757 dst1->next = dst_prev;
1758 dst_prev = dst1;
1759
1760 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1761 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1762 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1763 trailer_len += xfrm[i]->props.trailer_len;
1764 }
1765
1766 dst_prev->child = dst;
1767 dst0->path = dst;
1768
1769 err = -ENODEV;
1770 dev = dst->dev;
1771 if (!dev)
1772 goto free_dst;
1773
a1b05140 1774 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1775 xfrm_init_pmtu(dst_prev);
1776
1777 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1778 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1779
87c1e12b 1780 err = xfrm_fill_dst(xdst, dev, fl);
25ee3286
HX		 1781 if (err)
1782 goto free_dst;
1783
1784 dst_prev->header_len = header_len;
1785 dst_prev->trailer_len = trailer_len;
1786 header_len -= xdst->u.dst.xfrm->props.header_len;
1787 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1788 }
1789
1790out:
1791 return dst0;
1792
1793put_states:
1794 for (; i < nx; i++)
1795 xfrm_state_put(xfrm[i]);
1796free_dst:
1797 if (dst0)
52df157f 1798 dst_release_immediate(dst0);
25ee3286
HX		 1799 dst0 = ERR_PTR(err);
1800 goto out;
1801}
1802
73ff93cd 1803static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1804 struct xfrm_policy **pols,
1805 int *num_pols, int *num_xfrms)
1806{
1807 int i;
1808
1809 if (*num_pols == 0 || !pols[0]) {
1810 *num_pols = 0;
1811 *num_xfrms = 0;
1812 return 0;
1813 }
1814 if (IS_ERR(pols[0]))
1815 return PTR_ERR(pols[0]);
1816
1817 *num_xfrms = pols[0]->xfrm_nr;
1818
1819#ifdef CONFIG_XFRM_SUB_POLICY
1820 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1821 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1822 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1823 XFRM_POLICY_TYPE_MAIN,
1824 fl, family,
1825 XFRM_POLICY_OUT);
1826 if (pols[1]) {
1827 if (IS_ERR(pols[1])) {
1828 xfrm_pols_put(pols, *num_pols);
1829 return PTR_ERR(pols[1]);
1830 }
02d0892f 1831 (*num_pols)++;
80c802f3
TT		 1832 (*num_xfrms) += pols[1]->xfrm_nr;
1833 }
1834 }
1835#endif
1836 for (i = 0; i < *num_pols; i++) {
1837 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1838 *num_xfrms = -1;
1839 break;
1840 }
1841 }
1842
1843 return 0;
1844
1845}
1846
1847static struct xfrm_dst *
1848xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1849 const struct flowi *fl, u16 family,
80c802f3
TT		 1850 struct dst_entry *dst_orig)
1851{
1852 struct net *net = xp_net(pols[0]);
1853 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1854 struct dst_entry *dst;
1855 struct xfrm_dst *xdst;
1856 int err;
1857
1858 /* Try to instantiate a bundle */
1859 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
d809ec89
TT		 1860 if (err <= 0) {
1861 if (err != 0 && err != -EAGAIN)
80c802f3
TT		 1862 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1863 return ERR_PTR(err);
1864 }
1865
1866 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1867 if (IS_ERR(dst)) {
1868 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1869 return ERR_CAST(dst);
1870 }
1871
1872 xdst = (struct xfrm_dst *)dst;
1873 xdst->num_xfrms = err;
80c802f3 1874 xdst->num_pols = num_pols;
3e94c2dc 1875 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 1876 xdst->policy_genid = atomic_read(&pols[0]->genid);
1877
1878 return xdst;
1879}
1880
a0073fe1
SK		 1881static void xfrm_policy_queue_process(unsigned long arg)
1882{
a0073fe1
SK		 1883 struct sk_buff *skb;
1884 struct sock *sk;
1885 struct dst_entry *dst;
a0073fe1 1886 struct xfrm_policy *pol = (struct xfrm_policy *)arg;
3f5312ae 1887 struct net *net = xp_net(pol);
a0073fe1
SK		 1888 struct xfrm_policy_queue *pq = &pol->polq;
1889 struct flowi fl;
1890 struct sk_buff_head list;
1891
1892 spin_lock(&pq->hold_queue.lock);
1893 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK		 1894 if (!skb) {
1895 spin_unlock(&pq->hold_queue.lock);
1896 goto out;
1897 }
a0073fe1
SK		 1898 dst = skb_dst(skb);
1899 sk = skb->sk;
1900 xfrm_decode_session(skb, &fl, dst->ops->family);
1901 spin_unlock(&pq->hold_queue.lock);
1902
1903 dst_hold(dst->path);
3f5312ae 1904 dst = xfrm_lookup(net, dst->path, &fl, sk, 0);
a0073fe1
SK		 1905 if (IS_ERR(dst))
1906 goto purge_queue;
1907
1908 if (dst->flags & DST_XFRM_QUEUE) {
1909 dst_release(dst);
1910
1911 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1912 goto purge_queue;
1913
1914 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1915 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1916 xfrm_pol_hold(pol);
1917 goto out;
a0073fe1
SK		 1918 }
1919
1920 dst_release(dst);
1921
1922 __skb_queue_head_init(&list);
1923
1924 spin_lock(&pq->hold_queue.lock);
1925 pq->timeout = 0;
1926 skb_queue_splice_init(&pq->hold_queue, &list);
1927 spin_unlock(&pq->hold_queue.lock);
1928
1929 while (!skb_queue_empty(&list)) {
1930 skb = __skb_dequeue(&list);
1931
1932 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1933 dst_hold(skb_dst(skb)->path);
3f5312ae 1934 dst = xfrm_lookup(net, skb_dst(skb)->path, &fl, skb->sk, 0);
a0073fe1 1935 if (IS_ERR(dst)) {
a0073fe1
SK		 1936 kfree_skb(skb);
1937 continue;
1938 }
1939
1940 nf_reset(skb);
1941 skb_dst_drop(skb);
1942 skb_dst_set(skb, dst);
1943
13206b6b 1944 dst_output(net, skb->sk, skb);
a0073fe1
SK		 1945 }
1946
e7d8f6cb
SK		 1947out:
1948 xfrm_pol_put(pol);
a0073fe1
SK		 1949 return;
1950
1951purge_queue:
1952 pq->timeout = 0;
1ee5e667 1953 skb_queue_purge(&pq->hold_queue);
e7d8f6cb 1954 xfrm_pol_put(pol);
a0073fe1
SK		 1955}
1956
ede2059d 1957static int xdst_queue_output(struct net *net, struct sock *sk, struct sk_buff *skb)
a0073fe1
SK		 1958{
1959 unsigned long sched_next;
1960 struct dst_entry *dst = skb_dst(skb);
1961 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1962 struct xfrm_policy *pol = xdst->pols[0];
1963 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4 1964
39bb5e62 1965 if (unlikely(skb_fclone_busy(sk, skb))) {
4d53eff4
SK		 1966 kfree_skb(skb);
1967 return 0;
1968 }
a0073fe1
SK		 1969
1970 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1971 kfree_skb(skb);
1972 return -EAGAIN;
1973 }
1974
1975 skb_dst_force(skb);
a0073fe1
SK		 1976
1977 spin_lock_bh(&pq->hold_queue.lock);
1978
1979 if (!pq->timeout)
1980 pq->timeout = XFRM_QUEUE_TMO_MIN;
1981
1982 sched_next = jiffies + pq->timeout;
1983
1984 if (del_timer(&pq->hold_timer)) {
1985 if (time_before(pq->hold_timer.expires, sched_next))
1986 sched_next = pq->hold_timer.expires;
e7d8f6cb 1987 xfrm_pol_put(pol);
a0073fe1
SK		 1988 }
1989
1990 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1991 if (!mod_timer(&pq->hold_timer, sched_next))
1992 xfrm_pol_hold(pol);
a0073fe1
SK		 1993
1994 spin_unlock_bh(&pq->hold_queue.lock);
1995
1996 return 0;
1997}
1998
1999static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
b8c203b2 2000 struct xfrm_flo *xflo,
a0073fe1
SK		 2001 const struct flowi *fl,
2002 int num_xfrms,
2003 u16 family)
2004{
2005 int err;
2006 struct net_device *dev;
b8c203b2 2007 struct dst_entry *dst;
a0073fe1
SK		 2008 struct dst_entry *dst1;
2009 struct xfrm_dst *xdst;
2010
2011 xdst = xfrm_alloc_dst(net, family);
2012 if (IS_ERR(xdst))
2013 return xdst;
2014
b8c203b2
SK		 2015 if (!(xflo->flags & XFRM_LOOKUP_QUEUE) ||
2016 net->xfrm.sysctl_larval_drop ||
2017 num_xfrms <= 0)
a0073fe1
SK		 2018 return xdst;
2019
b8c203b2 2020 dst = xflo->dst_orig;
a0073fe1
SK		 2021 dst1 = &xdst->u.dst;
2022 dst_hold(dst);
2023 xdst->route = dst;
2024
2025 dst_copy_metrics(dst1, dst);
2026
2027 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
2028 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
2029 dst1->lastuse = jiffies;
2030
2031 dst1->input = dst_discard;
2032 dst1->output = xdst_queue_output;
2033
2034 dst_hold(dst);
2035 dst1->child = dst;
2036 dst1->path = dst;
2037
2038 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
2039
2040 err = -ENODEV;
2041 dev = dst->dev;
2042 if (!dev)
2043 goto free_dst;
2044
2045 err = xfrm_fill_dst(xdst, dev, fl);
2046 if (err)
2047 goto free_dst;
2048
2049out:
2050 return xdst;
2051
2052free_dst:
2053 dst_release(dst1);
2054 xdst = ERR_PTR(err);
2055 goto out;
2056}
2057
80c802f3 2058static struct flow_cache_object *
dee9f4bc 2059xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
80c802f3
TT		 2060 struct flow_cache_object *oldflo, void *ctx)
2061{
b8c203b2 2062 struct xfrm_flo *xflo = (struct xfrm_flo *)ctx;
80c802f3
TT		 2063 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2064 struct xfrm_dst *xdst, *new_xdst;
2065 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
2066
2067 /* Check if the policies from old bundle are usable */
2068 xdst = NULL;
2069 if (oldflo) {
2070 xdst = container_of(oldflo, struct xfrm_dst, flo);
2071 num_pols = xdst->num_pols;
2072 num_xfrms = xdst->num_xfrms;
2073 pol_dead = 0;
2074 for (i = 0; i < num_pols; i++) {
2075 pols[i] = xdst->pols[i];
2076 pol_dead |= pols[i]->walk.dead;
2077 }
2078 if (pol_dead) {
52df157f
WW		 2079 /* Mark DST_OBSOLETE_DEAD to fail the next
2080 * xfrm_dst_check()
2081 */
2082 xdst->u.dst.obsolete = DST_OBSOLETE_DEAD;
2083 dst_release_immediate(&xdst->u.dst);
80c802f3
TT		 2084 xdst = NULL;
2085 num_pols = 0;
2086 num_xfrms = 0;
2087 oldflo = NULL;
2088 }
2089 }
2090
2091 /* Resolve policies to use if we couldn't get them from
2092 * previous cache entry */
2093 if (xdst == NULL) {
2094 num_pols = 1;
b5fb82c4
BZ		 2095 pols[0] = __xfrm_policy_lookup(net, fl, family,
2096 flow_to_policy_dir(dir));
80c802f3
TT		 2097 err = xfrm_expand_policies(fl, family, pols,
2098 &num_pols, &num_xfrms);
2099 if (err < 0)
2100 goto inc_error;
2101 if (num_pols == 0)
2102 return NULL;
2103 if (num_xfrms <= 0)
2104 goto make_dummy_bundle;
2105 }
2106
b8c203b2
SK		 2107 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family,
2108 xflo->dst_orig);
80c802f3
TT		 2109 if (IS_ERR(new_xdst)) {
2110 err = PTR_ERR(new_xdst);
2111 if (err != -EAGAIN)
2112 goto error;
2113 if (oldflo == NULL)
2114 goto make_dummy_bundle;
2115 dst_hold(&xdst->u.dst);
2116 return oldflo;
d809ec89
TT		 2117 } else if (new_xdst == NULL) {
2118 num_xfrms = 0;
2119 if (oldflo == NULL)
2120 goto make_dummy_bundle;
2121 xdst->num_xfrms = 0;
2122 dst_hold(&xdst->u.dst);
2123 return oldflo;
80c802f3
TT		 2124 }
2125
2126 /* Kill the previous bundle */
2127 if (xdst) {
2128 /* The policies were stolen for newly generated bundle */
2129 xdst->num_pols = 0;
52df157f
WW		 2130 /* Mark DST_OBSOLETE_DEAD to fail the next xfrm_dst_check() */
2131 xdst->u.dst.obsolete = DST_OBSOLETE_DEAD;
2132 dst_release_immediate(&xdst->u.dst);
80c802f3
TT		 2133 }
2134
52df157f 2135 /* We do need to return one reference for original caller */
80c802f3
TT		 2136 dst_hold(&new_xdst->u.dst);
2137 return &new_xdst->flo;
2138
2139make_dummy_bundle:
2140 /* We found policies, but there's no bundles to instantiate:
2141 * either because the policy blocks, has no transformations or
2142 * we could not build template (no xfrm_states).*/
b8c203b2 2143 xdst = xfrm_create_dummy_bundle(net, xflo, fl, num_xfrms, family);
80c802f3
TT		 2144 if (IS_ERR(xdst)) {
2145 xfrm_pols_put(pols, num_pols);
2146 return ERR_CAST(xdst);
2147 }
2148 xdst->num_pols = num_pols;
2149 xdst->num_xfrms = num_xfrms;
3e94c2dc 2150 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2151
2152 dst_hold(&xdst->u.dst);
2153 return &xdst->flo;
2154
2155inc_error:
2156 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2157error:
52df157f
WW		 2158 if (xdst != NULL) {
2159 /* Mark DST_OBSOLETE_DEAD to fail the next xfrm_dst_check() */
2160 xdst->u.dst.obsolete = DST_OBSOLETE_DEAD;
2161 dst_release_immediate(&xdst->u.dst);
2162 } else
80c802f3
TT		 2163 xfrm_pols_put(pols, num_pols);
2164 return ERR_PTR(err);
2165}
1da177e4 2166
2774c131
DM		 2167static struct dst_entry *make_blackhole(struct net *net, u16 family,
2168 struct dst_entry *dst_orig)
2169{
37b10383 2170 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2774c131
DM		 2171 struct dst_entry *ret;
2172
2173 if (!afinfo) {
2174 dst_release(dst_orig);
433a1954 2175 return ERR_PTR(-EINVAL);
2774c131
DM		 2176 } else {
2177 ret = afinfo->blackhole_route(net, dst_orig);
2178 }
bdba9fe0 2179 rcu_read_unlock();
2774c131
DM		 2180
2181 return ret;
2182}
2183
1da177e4
LT		 2184/* Main function: finds/creates a bundle for given flow.
2185 *
2186 * At the moment we eat a raw IP route. Mostly to speed up lookups
2187 * on interfaces with disabled IPsec.
2188 */
452edd59
DM		 2189struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2190 const struct flowi *fl,
6f9c9615 2191 const struct sock *sk, int flags)
1da177e4 2192{
4e81bb83 2193 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3
TT		 2194 struct flow_cache_object *flo;
2195 struct xfrm_dst *xdst;
452edd59 2196 struct dst_entry *dst, *route;
80c802f3 2197 u16 family = dst_orig->ops->family;
df71837d 2198 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
4b021628 2199 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2200
80c802f3
TT		 2201 dst = NULL;
2202 xdst = NULL;
2203 route = NULL;
4e81bb83 2204
bd5eb35f 2205 sk = sk_const_to_full_sk(sk);
f7944fb1 2206 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3 2207 num_pols = 1;
4c86d777 2208 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl, family);
80c802f3
TT		 2209 err = xfrm_expand_policies(fl, family, pols,
2210 &num_pols, &num_xfrms);
2211 if (err < 0)
75b8c133 2212 goto dropdst;
80c802f3
TT		 2213
2214 if (num_pols) {
2215 if (num_xfrms <= 0) {
2216 drop_pols = num_pols;
2217 goto no_transform;
2218 }
2219
2220 xdst = xfrm_resolve_and_create_bundle(
2221 pols, num_pols, fl,
2222 family, dst_orig);
2223 if (IS_ERR(xdst)) {
2224 xfrm_pols_put(pols, num_pols);
2225 err = PTR_ERR(xdst);
2226 goto dropdst;
d809ec89
TT		 2227 } else if (xdst == NULL) {
2228 num_xfrms = 0;
2229 drop_pols = num_pols;
2230 goto no_transform;
80c802f3
TT		 2231 }
2232
b7eea454
SK		 2233 dst_hold(&xdst->u.dst);
2234 xdst->u.dst.flags |= DST_NOCACHE;
80c802f3 2235 route = xdst->route;
0aa64774 2236 }
3bccfbc7 2237 }
1da177e4 2238
80c802f3 2239 if (xdst == NULL) {
b8c203b2
SK		 2240 struct xfrm_flo xflo;
2241
2242 xflo.dst_orig = dst_orig;
2243 xflo.flags = flags;
2244
1da177e4 2245 /* To accelerate a bit... */
2518c7c2 2246 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2247 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2248 goto nopol;
1da177e4 2249
80c802f3 2250 flo = flow_cache_lookup(net, fl, family, dir,
b8c203b2 2251 xfrm_bundle_lookup, &xflo);
80c802f3
TT		 2252 if (flo == NULL)
2253 goto nopol;
fe1a5f03 2254 if (IS_ERR(flo)) {
80c802f3 2255 err = PTR_ERR(flo);
75b8c133 2256 goto dropdst;
d66e37a9 2257 }
80c802f3
TT		 2258 xdst = container_of(flo, struct xfrm_dst, flo);
2259
2260 num_pols = xdst->num_pols;
2261 num_xfrms = xdst->num_xfrms;
3e94c2dc 2262 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2263 route = xdst->route;
2264 }
2265
2266 dst = &xdst->u.dst;
2267 if (route == NULL && num_xfrms > 0) {
2268 /* The only case when xfrm_bundle_lookup() returns a
2269 * bundle with null route, is when the template could
2270 * not be resolved. It means policies are there, but
2271 * bundle could not be created, since we don't yet
2272 * have the xfrm_state's. We need to wait for KM to
2273 * negotiate new SA's or bail out with error.*/
2274 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2275 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
ac37e251
W		 2276 err = -EREMOTE;
2277 goto error;
80c802f3 2278 }
80c802f3 2279
5b8ef341 2280 err = -EAGAIN;
80c802f3
TT		 2281
2282 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2283 goto error;
1da177e4
LT		 2284 }
2285
80c802f3
TT		 2286no_transform:
2287 if (num_pols == 0)
8b7817f3 2288 goto nopol;
1da177e4 2289
80c802f3
TT		 2290 if ((flags & XFRM_LOOKUP_ICMP) &&
2291 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2292 err = -ENOENT;
8b7817f3 2293 goto error;
80c802f3 2294 }
8b7817f3 2295
80c802f3
TT		 2296 for (i = 0; i < num_pols; i++)
2297 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2298
80c802f3 2299 if (num_xfrms < 0) {
1da177e4 2300 /* Prohibit the flow */
59c9940e 2301 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2302 err = -EPERM;
2303 goto error;
80c802f3
TT		 2304 } else if (num_xfrms > 0) {
2305 /* Flow transformed */
80c802f3
TT		 2306 dst_release(dst_orig);
2307 } else {
2308 /* Flow passes untransformed */
2309 dst_release(dst);
452edd59 2310 dst = dst_orig;
1da177e4 2311 }
80c802f3
TT		 2312ok:
2313 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2314 if (dst && dst->xfrm &&
2315 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2316 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2317 return dst;
1da177e4 2318
80c802f3 2319nopol:
452edd59
DM		 2320 if (!(flags & XFRM_LOOKUP_ICMP)) {
2321 dst = dst_orig;
80c802f3 2322 goto ok;
452edd59 2323 }
80c802f3 2324 err = -ENOENT;
1da177e4 2325error:
80c802f3 2326 dst_release(dst);
75b8c133 2327dropdst:
ac37e251
W		 2328 if (!(flags & XFRM_LOOKUP_KEEP_DST_REF))
2329 dst_release(dst_orig);
80c802f3 2330 xfrm_pols_put(pols, drop_pols);
452edd59 2331 return ERR_PTR(err);
1da177e4
LT		 2332}
2333EXPORT_SYMBOL(xfrm_lookup);
2334
f92ee619
SK		 2335/* Callers of xfrm_lookup_route() must ensure a call to dst_output().
2336 * Otherwise we may send out blackholed packets.
2337 */
2338struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
2339 const struct flowi *fl,
6f9c9615 2340 const struct sock *sk, int flags)
f92ee619 2341{
b8c203b2 2342 struct dst_entry *dst = xfrm_lookup(net, dst_orig, fl, sk,
ac37e251
W		 2343 flags | XFRM_LOOKUP_QUEUE |
2344 XFRM_LOOKUP_KEEP_DST_REF);
f92ee619
SK		 2345
2346 if (IS_ERR(dst) && PTR_ERR(dst) == -EREMOTE)
2347 return make_blackhole(net, dst_orig->ops->family, dst_orig);
2348
2349 return dst;
2350}
2351EXPORT_SYMBOL(xfrm_lookup_route);
2352
df0ba92a 2353static inline int
8f029de2 2354xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2355{
2356 struct xfrm_state *x;
df0ba92a
MN		 2357
2358 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2359 return 0;
2360 x = skb->sp->xvec[idx];
2361 if (!x->type->reject)
2362 return 0;
1ecafede 2363 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2364}
2365
1da177e4
LT		 2366/* When skb is transformed back to its "native" form, we have to
2367 * check policy restrictions. At the moment we make this in maximally
2368 * stupid way. Shame on me. :-) Of course, connected sockets must
2369 * have policy cached at them.
2370 */
2371
2372static inline int
7db454b9 2373xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2374 unsigned short family)
2375{
2376 if (xfrm_state_kern(x))
928ba416 2377 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2378 return x->id.proto == tmpl->id.proto &&
2379 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2380 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2381 x->props.mode == tmpl->mode &&
c5d18e98 2382 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2383 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2384 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2385 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2386}
2387
df0ba92a
MN		 2388/*
2389 * 0 or more than 0 is returned when validation is succeeded (either bypass
2390 * because of optional transport mode, or next index of the mathced secpath
2391 * state with the template.
2392 * -1 is returned when no matching template is found.
2393 * Otherwise "-2 - errored_index" is returned.
2394 */
1da177e4 2395static inline int
22cccb7e 2396xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2397 unsigned short family)
2398{
2399 int idx = start;
2400
2401 if (tmpl->optional) {
7e49e6de 2402 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2403 return start;
2404 } else
2405 start = -1;
2406 for (; idx < sp->len; idx++) {
dbe5b4aa 2407 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2408 return ++idx;
df0ba92a
MN		 2409 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2410 if (start == -1)
2411 start = -2-idx;
1da177e4 2412 break;
df0ba92a 2413 }
1da177e4
LT		 2414 }
2415 return start;
2416}
2417
d5422efe
HX		 2418int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2419 unsigned int family, int reverse)
1da177e4 2420{
37b10383 2421 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2422 int err;
1da177e4
LT		 2423
2424 if (unlikely(afinfo == NULL))
2425 return -EAFNOSUPPORT;
2426
d5422efe 2427 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2428 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
bdba9fe0 2429 rcu_read_unlock();
e0d1caa7 2430 return err;
1da177e4 2431}
d5422efe 2432EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2433
9a7386ec 2434static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2435{
2436 for (; k < sp->len; k++) {
df0ba92a 2437 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2438 *idxp = k;
1da177e4 2439 return 1;
df0ba92a 2440 }
1da177e4
LT		 2441 }
2442
2443 return 0;
2444}
2445
a716c119 2446int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2447 unsigned short family)
2448{
f6e1e25d 2449 struct net *net = dev_net(skb->dev);
1da177e4 2450 struct xfrm_policy *pol;
4e81bb83
MN		 2451 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2452 int npols = 0;
2453 int xfrm_nr;
2454 int pi;
d5422efe 2455 int reverse;
1da177e4 2456 struct flowi fl;
d5422efe 2457 u8 fl_dir;
df0ba92a 2458 int xerr_idx = -1;
1da177e4 2459
d5422efe
HX		 2460 reverse = dir & ~XFRM_POLICY_MASK;
2461 dir &= XFRM_POLICY_MASK;
2462 fl_dir = policy_to_flow_dir(dir);
2463
0aa64774 2464 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2465 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2466 return 0;
0aa64774
MN		 2467 }
2468
eb9c7ebe 2469 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2470
2471 /* First, check used SA against their selectors. */
2472 if (skb->sp) {
2473 int i;
2474
9b7a787d 2475 for (i = skb->sp->len-1; i >= 0; i--) {
dbe5b4aa 2476 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2477 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2478 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2479 return 0;
0aa64774 2480 }
1da177e4
LT		 2481 }
2482 }
2483
2484 pol = NULL;
bd5eb35f 2485 sk = sk_to_full_sk(sk);
3bccfbc7 2486 if (sk && sk->sk_policy[dir]) {
4c86d777 2487 pol = xfrm_sk_policy_lookup(sk, dir, &fl, family);
0aa64774 2488 if (IS_ERR(pol)) {
59c9940e 2489 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2490 return 0;
0aa64774 2491 }
3bccfbc7 2492 }
1da177e4 2493
fe1a5f03
TT		 2494 if (!pol) {
2495 struct flow_cache_object *flo;
2496
2497 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2498 xfrm_policy_lookup, NULL);
2499 if (IS_ERR_OR_NULL(flo))
2500 pol = ERR_CAST(flo);
2501 else
2502 pol = container_of(flo, struct xfrm_policy, flo);
2503 }
1da177e4 2504
0aa64774 2505 if (IS_ERR(pol)) {
59c9940e 2506 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2507 return 0;
0aa64774 2508 }
134b0fc5 2509
df0ba92a 2510 if (!pol) {
d1d9facf 2511 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2512 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2513 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2514 return 0;
2515 }
2516 return 1;
2517 }
1da177e4 2518
9d729f72 2519 pol->curlft.use_time = get_seconds();
1da177e4 2520
4e81bb83 2521 pols[0] = pol;
02d0892f 2522 npols++;
4e81bb83
MN		 2523#ifdef CONFIG_XFRM_SUB_POLICY
2524 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2525 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2526 &fl, family,
2527 XFRM_POLICY_IN);
2528 if (pols[1]) {
0aa64774 2529 if (IS_ERR(pols[1])) {
59c9940e 2530 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2531 return 0;
0aa64774 2532 }
9d729f72 2533 pols[1]->curlft.use_time = get_seconds();
02d0892f 2534 npols++;
4e81bb83
MN		 2535 }
2536 }
2537#endif
2538
1da177e4
LT		 2539 if (pol->action == XFRM_POLICY_ALLOW) {
2540 struct sec_path *sp;
2541 static struct sec_path dummy;
4e81bb83 2542 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2543 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2544 struct xfrm_tmpl **tpp = tp;
2545 int ti = 0;
1da177e4
LT		 2546 int i, k;
2547
2548 if ((sp = skb->sp) == NULL)
2549 sp = &dummy;
2550
4e81bb83
MN		 2551 for (pi = 0; pi < npols; pi++) {
2552 if (pols[pi] != pol &&
0aa64774 2553 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2554 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2555 goto reject;
0aa64774
MN		 2556 }
2557 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2558 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2559 goto reject_error;
0aa64774 2560 }
4e81bb83
MN		 2561 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2562 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2563 }
2564 xfrm_nr = ti;
41a49cc3 2565 if (npols > 1) {
283bc9f3 2566 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family, net);
41a49cc3
MN		 2567 tpp = stp;
2568 }
4e81bb83 2569
1da177e4
LT		 2570 /* For each tunnel xfrm, find the first matching tmpl.
2571 * For each tmpl before that, find corresponding xfrm.
2572 * Order is _important_. Later we will implement
2573 * some barriers, but at the moment barriers
2574 * are implied between each two transformations.
2575 */
4e81bb83
MN		 2576 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2577 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2578 if (k < 0) {
d1d9facf
JM		 2579 if (k < -1)
2580 /* "-2 - errored_index" returned */
2581 xerr_idx = -(2+k);
59c9940e 2582 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2583 goto reject;
df0ba92a 2584 }
1da177e4
LT		 2585 }
2586
0aa64774 2587 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2588 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2589 goto reject;
0aa64774 2590 }
1da177e4 2591
4e81bb83 2592 xfrm_pols_put(pols, npols);
1da177e4
LT		 2593 return 1;
2594 }
59c9940e 2595 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2596
2597reject:
df0ba92a 2598 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2599reject_error:
2600 xfrm_pols_put(pols, npols);
1da177e4
LT		 2601 return 0;
2602}
2603EXPORT_SYMBOL(__xfrm_policy_check);
2604
2605int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2606{
99a66657 2607 struct net *net = dev_net(skb->dev);
1da177e4 2608 struct flowi fl;
adf30907 2609 struct dst_entry *dst;
73137147 2610 int res = 1;
1da177e4 2611
0aa64774 2612 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2613 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2614 return 0;
0aa64774 2615 }
1da177e4 2616
fafeeb6c 2617 skb_dst_force(skb);
adf30907 2618
b8c203b2 2619 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, XFRM_LOOKUP_QUEUE);
452edd59 2620 if (IS_ERR(dst)) {
73137147 2621 res = 0;
452edd59
DM		 2622 dst = NULL;
2623 }
adf30907
ED		 2624 skb_dst_set(skb, dst);
2625 return res;
1da177e4
LT		 2626}
2627EXPORT_SYMBOL(__xfrm_route_forward);
2628
d49c73c7
DM		 2629/* Optimize later using cookies and generation ids. */
2630
1da177e4
LT		 2631static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2632{
d49c73c7 2633 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2634 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2635 * get validated by dst_ops->check on every use. We do this
2636 * because when a normal route referenced by an XFRM dst is
2637 * obsoleted we do not go looking around for all parent
2638 * referencing XFRM dsts so that we can invalidate them. It
2639 * is just too much work. Instead we make the checks here on
2640 * every use. For example:
d49c73c7
DM		 2641 *
2642 * XFRM dst A --> IPv4 dst X
2643 *
2644 * X is the "xdst->route" of A (X is also the "dst->path" of A
2645 * in this example). If X is marked obsolete, "A" will not
2646 * notice. That's what we are validating here via the
2647 * stale_bundle() check.
2648 *
52df157f
WW		 2649 * When an xdst is removed from flow cache, DST_OBSOLETE_DEAD will
2650 * be marked on it.
2651 * When a dst is removed from the fib tree, DST_OBSOLETE_DEAD will
2652 * be marked on it.
2653 * Both will force stable_bundle() to fail on any xdst bundle with
2654 * this dst linked in it.
399c180a 2655 */
d49c73c7
DM		 2656 if (dst->obsolete < 0 && !stale_bundle(dst))
2657 return dst;
2658
1da177e4
LT		 2659 return NULL;
2660}
2661
2662static int stale_bundle(struct dst_entry *dst)
2663{
12fdb4d3 2664 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2665}
2666
aabc9761 2667void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2668{
1da177e4 2669 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2670 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2671 dev_hold(dst->dev);
1da177e4
LT		 2672 dev_put(dev);
2673 }
2674}
aabc9761 2675EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2676
2677static void xfrm_link_failure(struct sk_buff *skb)
2678{
2679 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2680}
2681
2682static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2683{
2684 if (dst) {
2685 if (dst->obsolete) {
2686 dst_release(dst);
2687 dst = NULL;
2688 }
2689 }
2690 return dst;
2691}
2692
e4c17216 2693void xfrm_garbage_collect(struct net *net)
c0ed1c14 2694{
ca925cf1 2695 flow_cache_flush(net);
c0ed1c14 2696}
e4c17216 2697EXPORT_SYMBOL(xfrm_garbage_collect);
c0ed1c14 2698
3d7d25a6 2699void xfrm_garbage_collect_deferred(struct net *net)
c0ed1c14 2700{
ca925cf1 2701 flow_cache_flush_deferred(net);
c0ed1c14 2702}
3d7d25a6 2703EXPORT_SYMBOL(xfrm_garbage_collect_deferred);
c0ed1c14 2704
25ee3286 2705static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2706{
2707 do {
2708 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2709 u32 pmtu, route_mtu_cached;
2710
2711 pmtu = dst_mtu(dst->child);
2712 xdst->child_mtu_cached = pmtu;
2713
2714 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2715
2716 route_mtu_cached = dst_mtu(xdst->route);
2717 xdst->route_mtu_cached = route_mtu_cached;
2718
2719 if (pmtu > route_mtu_cached)
2720 pmtu = route_mtu_cached;
2721
defb3519 2722 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2723 } while ((dst = dst->next));
2724}
2725
1da177e4
LT		 2726/* Check that the bundle accepts the flow and its components are
2727 * still valid.
2728 */
2729
12fdb4d3 2730static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2731{
2732 struct dst_entry *dst = &first->u.dst;
2733 struct xfrm_dst *last;
2734 u32 mtu;
2735
92d63dec 2736 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2737 (dst->dev && !netif_running(dst->dev)))
2738 return 0;
2739
a0073fe1
SK		 2740 if (dst->flags & DST_XFRM_QUEUE)
2741 return 1;
2742
1da177e4
LT		 2743 last = NULL;
2744
2745 do {
2746 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2747
1da177e4
LT		 2748 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2749 return 0;
80c802f3
TT		 2750 if (xdst->xfrm_genid != dst->xfrm->genid)
2751 return 0;
b1312c89
TT		 2752 if (xdst->num_pols > 0 &&
2753 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2754 return 0;
e53820de 2755
1da177e4
LT		 2756 mtu = dst_mtu(dst->child);
2757 if (xdst->child_mtu_cached != mtu) {
2758 last = xdst;
2759 xdst->child_mtu_cached = mtu;
2760 }
2761
92d63dec 2762 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2763 return 0;
2764 mtu = dst_mtu(xdst->route);
2765 if (xdst->route_mtu_cached != mtu) {
2766 last = xdst;
2767 xdst->route_mtu_cached = mtu;
2768 }
2769
2770 dst = dst->child;
2771 } while (dst->xfrm);
2772
2773 if (likely(!last))
2774 return 1;
2775
2776 mtu = last->child_mtu_cached;
2777 for (;;) {
2778 dst = &last->u.dst;
2779
2780 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2781 if (mtu > last->route_mtu_cached)
2782 mtu = last->route_mtu_cached;
defb3519 2783 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2784
2785 if (last == first)
2786 break;
2787
bd0bf076 2788 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2789 last->child_mtu_cached = mtu;
2790 }
2791
2792 return 1;
2793}
2794
0dbaee3b
DM		 2795static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2796{
2797 return dst_metric_advmss(dst->path);
2798}
2799
ebb762f2 2800static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2801{
618f9bc7
SK		 2802 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2803
2804 return mtu ? : dst_mtu(dst->path);
d33e4553
DM		 2805}
2806
1ecc9ad0
JA		 2807static const void *xfrm_get_dst_nexthop(const struct dst_entry *dst,
2808 const void *daddr)
63fca65d
JA		 2809{
2810 const struct dst_entry *path = dst->path;
2811
2812 for (; dst != path; dst = dst->child) {
2813 const struct xfrm_state *xfrm = dst->xfrm;
2814
2815 if (xfrm->props.mode == XFRM_MODE_TRANSPORT)
2816 continue;
2817 if (xfrm->type->flags & XFRM_TYPE_REMOTE_COADDR)
2818 daddr = xfrm->coaddr;
2819 else if (!(xfrm->type->flags & XFRM_TYPE_LOCAL_COADDR))
2820 daddr = &xfrm->id.daddr;
2821 }
1ecc9ad0
JA		 2822 return daddr;
2823}
2824
2825static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2826 struct sk_buff *skb,
2827 const void *daddr)
2828{
2829 const struct dst_entry *path = dst->path;
2830
2831 if (!skb)
2832 daddr = xfrm_get_dst_nexthop(dst, daddr);
2833 return path->ops->neigh_lookup(path, skb, daddr);
2834}
2835
2836static void xfrm_confirm_neigh(const struct dst_entry *dst, const void *daddr)
2837{
2838 const struct dst_entry *path = dst->path;
2839
2840 daddr = xfrm_get_dst_nexthop(dst, daddr);
63fca65d
JA		 2841 path->ops->confirm_neigh(path, daddr);
2842}
2843
a2817d8b 2844int xfrm_policy_register_afinfo(const struct xfrm_policy_afinfo *afinfo, int family)
1da177e4
LT		 2845{
2846 int err = 0;
a2817d8b
FW		 2847
2848 if (WARN_ON(family >= ARRAY_SIZE(xfrm_policy_afinfo)))
1da177e4 2849 return -EAFNOSUPPORT;
a2817d8b 2850
ef8531b6 2851 spin_lock(&xfrm_policy_afinfo_lock);
a2817d8b 2852 if (unlikely(xfrm_policy_afinfo[family] != NULL))
f31e8d4f 2853 err = -EEXIST;
1da177e4
LT		 2854 else {
2855 struct dst_ops *dst_ops = afinfo->dst_ops;
2856 if (likely(dst_ops->kmem_cachep == NULL))
2857 dst_ops->kmem_cachep = xfrm_dst_cache;
2858 if (likely(dst_ops->check == NULL))
2859 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2860 if (likely(dst_ops->default_advmss == NULL))
2861 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2862 if (likely(dst_ops->mtu == NULL))
2863 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2864 if (likely(dst_ops->negative_advice == NULL))
2865 dst_ops->negative_advice = xfrm_negative_advice;
2866 if (likely(dst_ops->link_failure == NULL))
2867 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2868 if (likely(dst_ops->neigh_lookup == NULL))
2869 dst_ops->neigh_lookup = xfrm_neigh_lookup;
63fca65d
JA		 2870 if (likely(!dst_ops->confirm_neigh))
2871 dst_ops->confirm_neigh = xfrm_confirm_neigh;
a2817d8b 2872 rcu_assign_pointer(xfrm_policy_afinfo[family], afinfo);
1da177e4 2873 }
ef8531b6 2874 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c 2875
1da177e4
LT		 2876 return err;
2877}
2878EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2879
a2817d8b 2880void xfrm_policy_unregister_afinfo(const struct xfrm_policy_afinfo *afinfo)
1da177e4 2881{
2b61997a 2882 struct dst_ops *dst_ops = afinfo->dst_ops;
a2817d8b 2883 int i;
2b61997a 2884
a2817d8b
FW		 2885 for (i = 0; i < ARRAY_SIZE(xfrm_policy_afinfo); i++) {
2886 if (xfrm_policy_afinfo[i] != afinfo)
2887 continue;
2888 RCU_INIT_POINTER(xfrm_policy_afinfo[i], NULL);
2889 break;
ef8531b6 2890 }
ef8531b6 2891
2b61997a 2892 synchronize_rcu();
ef8531b6 2893
2b61997a
FW		 2894 dst_ops->kmem_cachep = NULL;
2895 dst_ops->check = NULL;
2896 dst_ops->negative_advice = NULL;
2897 dst_ops->link_failure = NULL;
1da177e4
LT		 2898}
2899EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2900
558f82ef 2901#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2902static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2903{
c68cd1a0 2904 int rv;
698365fa
WC		 2905 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib);
2906 if (!net->mib.xfrm_statistics)
558f82ef 2907 return -ENOMEM;
c68cd1a0
AD		 2908 rv = xfrm_proc_init(net);
2909 if (rv < 0)
698365fa 2910 free_percpu(net->mib.xfrm_statistics);
c68cd1a0 2911 return rv;
558f82ef 2912}
59c9940e
AD		 2913
2914static void xfrm_statistics_fini(struct net *net)
2915{
c68cd1a0 2916 xfrm_proc_fini(net);
698365fa 2917 free_percpu(net->mib.xfrm_statistics);
59c9940e
AD		 2918}
2919#else
2920static int __net_init xfrm_statistics_init(struct net *net)
2921{
2922 return 0;
2923}
2924
2925static void xfrm_statistics_fini(struct net *net)
2926{
2927}
558f82ef
MN		 2928#endif
2929
d62ddc21 2930static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2931{
2518c7c2
DM		 2932 unsigned int hmask, sz;
2933 int dir;
2934
d62ddc21
AD		 2935 if (net_eq(net, &init_net))
2936 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2937 sizeof(struct xfrm_dst),
e5d679f3 2938 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2939 NULL);
1da177e4 2940
2518c7c2
DM		 2941 hmask = 8 - 1;
2942 sz = (hmask+1) * sizeof(struct hlist_head);
2943
93b851c1
AD		 2944 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2945 if (!net->xfrm.policy_byidx)
2946 goto out_byidx;
8100bea7 2947 net->xfrm.policy_idx_hmask = hmask;
2518c7c2 2948
53c2e285 2949 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2
DM		 2950 struct xfrm_policy_hash *htab;
2951
dc2caba7 2952 net->xfrm.policy_count[dir] = 0;
53c2e285 2953 net->xfrm.policy_count[XFRM_POLICY_MAX + dir] = 0;
8b18f8ea 2954 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2955
a35f6c5d 2956 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2957 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2958 if (!htab->table)
a35f6c5d
AD		 2959 goto out_bydst;
2960 htab->hmask = hmask;
b58555f1
CG		 2961 htab->dbits4 = 32;
2962 htab->sbits4 = 32;
2963 htab->dbits6 = 128;
2964 htab->sbits6 = 128;
2518c7c2 2965 }
880a6fab
CG		 2966 net->xfrm.policy_hthresh.lbits4 = 32;
2967 net->xfrm.policy_hthresh.rbits4 = 32;
2968 net->xfrm.policy_hthresh.lbits6 = 128;
2969 net->xfrm.policy_hthresh.rbits6 = 128;
2970
2971 seqlock_init(&net->xfrm.policy_hthresh.lock);
2518c7c2 2972
adfcf0b2 2973 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2974 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
880a6fab 2975 INIT_WORK(&net->xfrm.policy_hthresh.work, xfrm_hash_rebuild);
d62ddc21 2976 if (net_eq(net, &init_net))
21f42cc9 2977 xfrm_dev_init();
d62ddc21 2978 return 0;
93b851c1 2979
a35f6c5d
AD		 2980out_bydst:
2981 for (dir--; dir >= 0; dir--) {
2982 struct xfrm_policy_hash *htab;
2983
2984 htab = &net->xfrm.policy_bydst[dir];
2985 xfrm_hash_free(htab->table, sz);
2986 }
2987 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2988out_byidx:
2989 return -ENOMEM;
d62ddc21
AD		 2990}
2991
2992static void xfrm_policy_fini(struct net *net)
2993{
93b851c1 2994 unsigned int sz;
8b18f8ea 2995 int dir;
93b851c1 2996
7c2776ee
AD		 2997 flush_work(&net->xfrm.policy_hash_work);
2998#ifdef CONFIG_XFRM_SUB_POLICY
2e71029e 2999 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false);
7c2776ee 3000#endif
2e71029e 3001 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false);
7c2776ee 3002
adfcf0b2 3003 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 3004
53c2e285 3005 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
a35f6c5d
AD		 3006 struct xfrm_policy_hash *htab;
3007
8b18f8ea 3008 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 3009
3010 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 3011 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 3012 WARN_ON(!hlist_empty(htab->table));
3013 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 3014 }
3015
8100bea7 3016 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 3017 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
3018 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 3019}
3020
d62ddc21
AD		 3021static int __net_init xfrm_net_init(struct net *net)
3022{
3023 int rv;
3024
c282222a
FW		 3025 /* Initialize the per-net locks here */
3026 spin_lock_init(&net->xfrm.xfrm_state_lock);
3027 spin_lock_init(&net->xfrm.xfrm_policy_lock);
3028 mutex_init(&net->xfrm.xfrm_cfg_mutex);
3029
59c9940e
AD		 3030 rv = xfrm_statistics_init(net);
3031 if (rv < 0)
3032 goto out_statistics;
d62ddc21
AD		 3033 rv = xfrm_state_init(net);
3034 if (rv < 0)
3035 goto out_state;
3036 rv = xfrm_policy_init(net);
3037 if (rv < 0)
3038 goto out_policy;
b27aeadb
AD		 3039 rv = xfrm_sysctl_init(net);
3040 if (rv < 0)
3041 goto out_sysctl;
4a93f509
SK		 3042 rv = flow_cache_init(net);
3043 if (rv < 0)
3044 goto out;
283bc9f3 3045
d62ddc21
AD		 3046 return 0;
3047
4a93f509
SK		 3048out:
3049 xfrm_sysctl_fini(net);
b27aeadb
AD		 3050out_sysctl:
3051 xfrm_policy_fini(net);
d62ddc21
AD		 3052out_policy:
3053 xfrm_state_fini(net);
3054out_state:
59c9940e
AD		 3055 xfrm_statistics_fini(net);
3056out_statistics:
d62ddc21
AD		 3057 return rv;
3058}
3059
3060static void __net_exit xfrm_net_exit(struct net *net)
3061{
4a93f509 3062 flow_cache_fini(net);
b27aeadb 3063 xfrm_sysctl_fini(net);
d62ddc21
AD		 3064 xfrm_policy_fini(net);
3065 xfrm_state_fini(net);
59c9940e 3066 xfrm_statistics_fini(net);
d62ddc21
AD		 3067}
3068
3069static struct pernet_operations __net_initdata xfrm_net_ops = {
3070 .init = xfrm_net_init,
3071 .exit = xfrm_net_exit,
3072};
3073
1da177e4
LT		 3074void __init xfrm_init(void)
3075{
a4fc1bfc 3076 flow_cache_hp_init();
d62ddc21 3077 register_pernet_subsys(&xfrm_net_ops);
30846090 3078 seqcount_init(&xfrm_policy_hash_generation);
1da177e4
LT		 3079 xfrm_input_init();
3080}
3081
ab5f5e8b 3082#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 3083static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
3084 struct audit_buffer *audit_buf)
ab5f5e8b 3085{
875179fa
PM		 3086 struct xfrm_sec_ctx *ctx = xp->security;
3087 struct xfrm_selector *sel = &xp->selector;
3088
3089 if (ctx)
ab5f5e8b 3090 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 3091 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 3092
9b7a787d 3093 switch (sel->family) {
ab5f5e8b 3094 case AF_INET:
21454aaa 3095 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 3096 if (sel->prefixlen_s != 32)
3097 audit_log_format(audit_buf, " src_prefixlen=%d",
3098 sel->prefixlen_s);
21454aaa 3099 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 3100 if (sel->prefixlen_d != 32)
3101 audit_log_format(audit_buf, " dst_prefixlen=%d",
3102 sel->prefixlen_d);
ab5f5e8b
JL		 3103 break;
3104 case AF_INET6:
5b095d98 3105 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 3106 if (sel->prefixlen_s != 128)
3107 audit_log_format(audit_buf, " src_prefixlen=%d",
3108 sel->prefixlen_s);
5b095d98 3109 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3110 if (sel->prefixlen_d != 128)
3111 audit_log_format(audit_buf, " dst_prefixlen=%d",
3112 sel->prefixlen_d);
ab5f5e8b
JL		 3113 break;
3114 }
3115}
3116
2e71029e 3117void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid)
ab5f5e8b
JL		 3118{
3119 struct audit_buffer *audit_buf;
ab5f5e8b 3120
afeb14b4 3121 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3122 if (audit_buf == NULL)
3123 return;
2e71029e 3124 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3125 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3126 xfrm_audit_common_policyinfo(xp, audit_buf);
3127 audit_log_end(audit_buf);
3128}
3129EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3130
68277acc 3131void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2e71029e 3132 bool task_valid)
ab5f5e8b
JL		 3133{
3134 struct audit_buffer *audit_buf;
ab5f5e8b 3135
afeb14b4 3136 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3137 if (audit_buf == NULL)
3138 return;
2e71029e 3139 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3140 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3141 xfrm_audit_common_policyinfo(xp, audit_buf);
3142 audit_log_end(audit_buf);
3143}
3144EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3145#endif
3146
80c9abaa 3147#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3148static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3149 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3150{
3151 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3152 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3153 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3154 sel_cmp->family) &&
3155 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3156 sel_cmp->family) &&
80c9abaa
SS		 3157 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3158 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3159 return true;
80c9abaa
SS		 3160 }
3161 } else {
3162 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3163 return true;
80c9abaa
SS		 3164 }
3165 }
bc9b35ad 3166 return false;
80c9abaa
SS		 3167}
3168
3e94c2dc
WC		 3169static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel,
3170 u8 dir, u8 type, struct net *net)
80c9abaa
SS		 3171{
3172 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3173 struct hlist_head *chain;
3174 u32 priority = ~0U;
3175
9d0380df 3176 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d549c4f 3177 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3178 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3179 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3180 pol->type == type) {
3181 ret = pol;
3182 priority = ret->priority;
3183 break;
3184 }
3185 }
8d549c4f 3186 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3187 hlist_for_each_entry(pol, chain, bydst) {
8faf491e
LR		 3188 if ((pol->priority >= priority) && ret)
3189 break;
3190
80c9abaa 3191 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
8faf491e 3192 pol->type == type) {
80c9abaa
SS		 3193 ret = pol;
3194 break;
3195 }
3196 }
3197
586f2eb4 3198 xfrm_pol_hold(ret);
80c9abaa 3199
9d0380df 3200 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
80c9abaa
SS		 3201
3202 return ret;
3203}
3204
dd701754 3205static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3206{
3207 int match = 0;
3208
3209 if (t->mode == m->mode && t->id.proto == m->proto &&
3210 (m->reqid == 0 || t->reqid == m->reqid)) {
3211 switch (t->mode) {
3212 case XFRM_MODE_TUNNEL:
3213 case XFRM_MODE_BEET:
70e94e66
YH		 3214 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3215 m->old_family) &&
3216 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3217 m->old_family)) {
80c9abaa
SS		 3218 match = 1;
3219 }
3220 break;
3221 case XFRM_MODE_TRANSPORT:
3222 /* in case of transport mode, template does not store
3223 any IP addresses, hence we just compare mode and
3224 protocol */
3225 match = 1;
3226 break;
3227 default:
3228 break;
3229 }
3230 }
3231 return match;
3232}
3233
3234/* update endpoint address(es) of template(s) */
3235static int xfrm_policy_migrate(struct xfrm_policy *pol,
3236 struct xfrm_migrate *m, int num_migrate)
3237{
3238 struct xfrm_migrate *mp;
80c9abaa
SS		 3239 int i, j, n = 0;
3240
3241 write_lock_bh(&pol->lock);
12a169e7 3242 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3243 /* target policy has been deleted */
3244 write_unlock_bh(&pol->lock);
3245 return -ENOENT;
3246 }
3247
3248 for (i = 0; i < pol->xfrm_nr; i++) {
3249 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3250 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3251 continue;
3252 n++;
1bfcb10f
HX		 3253 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3254 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3255 continue;
3256 /* update endpoints */
3257 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3258 sizeof(pol->xfrm_vec[i].id.daddr));
3259 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3260 sizeof(pol->xfrm_vec[i].saddr));
3261 pol->xfrm_vec[i].encap_family = mp->new_family;
3262 /* flush bundles */
80c802f3 3263 atomic_inc(&pol->genid);
80c9abaa
SS		 3264 }
3265 }
3266
3267 write_unlock_bh(&pol->lock);
3268
3269 if (!n)
3270 return -ENODATA;
3271
3272 return 0;
3273}
3274
dd701754 3275static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3276{
3277 int i, j;
3278
3279 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3280 return -EINVAL;
3281
3282 for (i = 0; i < num_migrate; i++) {
70e94e66
YH		 3283 if (xfrm_addr_equal(&m[i].old_daddr, &m[i].new_daddr,
3284 m[i].old_family) &&
3285 xfrm_addr_equal(&m[i].old_saddr, &m[i].new_saddr,
3286 m[i].old_family))
80c9abaa
SS		 3287 return -EINVAL;
3288 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3289 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3290 return -EINVAL;
3291
3292 /* check if there is any duplicated entry */
3293 for (j = i + 1; j < num_migrate; j++) {
3294 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3295 sizeof(m[i].old_daddr)) &&
3296 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3297 sizeof(m[i].old_saddr)) &&
3298 m[i].proto == m[j].proto &&
3299 m[i].mode == m[j].mode &&
3300 m[i].reqid == m[j].reqid &&
3301 m[i].old_family == m[j].old_family)
3302 return -EINVAL;
3303 }
3304 }
3305
3306 return 0;
3307}
3308
b4b7c0b3 3309int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3310 struct xfrm_migrate *m, int num_migrate,
8d549c4f 3311 struct xfrm_kmaddress *k, struct net *net)
80c9abaa
SS		 3312{
3313 int i, err, nx_cur = 0, nx_new = 0;
3314 struct xfrm_policy *pol = NULL;
3315 struct xfrm_state *x, *xc;
3316 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3317 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3318 struct xfrm_migrate *mp;
3319
3320 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3321 goto out;
3322
3323 /* Stage 1 - find policy */
8d549c4f 3324 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS		 3325 err = -ENOENT;
3326 goto out;
3327 }
3328
3329 /* Stage 2 - find and update state(s) */
3330 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
283bc9f3 3331 if ((x = xfrm_migrate_state_find(mp, net))) {
80c9abaa
SS		 3332 x_cur[nx_cur] = x;
3333 nx_cur++;
3334 if ((xc = xfrm_state_migrate(x, mp))) {
3335 x_new[nx_new] = xc;
3336 nx_new++;
3337 } else {
3338 err = -ENODATA;
3339 goto restore_state;
3340 }
3341 }
3342 }
3343
3344 /* Stage 3 - update policy */
3345 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3346 goto restore_state;
3347
3348 /* Stage 4 - delete old state(s) */
3349 if (nx_cur) {
3350 xfrm_states_put(x_cur, nx_cur);
3351 xfrm_states_delete(x_cur, nx_cur);
3352 }
3353
3354 /* Stage 5 - announce */
13c1d189 3355 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 3356
3357 xfrm_pol_put(pol);
3358
3359 return 0;
3360out:
3361 return err;
3362
3363restore_state:
3364 if (pol)
3365 xfrm_pol_put(pol);
3366 if (nx_cur)
3367 xfrm_states_put(x_cur, nx_cur);
3368 if (nx_new)
3369 xfrm_states_delete(x_new, nx_new);
3370
3371 return err;
3372}
e610e679 3373EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3374#endif
Ubuntu Artful kernel mirror